WEBVTT
1
00:00:00.120 --> 00:00:10.080
Welcome to the Data Strategy Gurus podcast. In this show, we bring together the brightest minds in the world of data strategy, data management, artificial intelligence, and disruptive technologies.
2
00:00:10.320 --> 00:00:16.860
Thought leaders and experts share their insights, knowledge, and experience on how to stay ahead of the game in an ever-evolving data landscape.
3
00:00:17.160 --> 00:00:23.540
Whether you're a data professional, a business leader, or simply someone who is passionate about the power of data, this podcast is for you.
4
00:00:23.820 --> 00:00:35.180
So sit back, relax, and join us on a journey to explore the world of data, analytics, artificial intelligence, tech, and beyond. Hi, and welcome to the Data Strategy Gurus podcast.
5
00:00:35.580 --> 00:00:49.340
F- today we have the VP of technology, Mathieu Jamblé, a former MarkLogic, but now integrated with the Progress data platform. Mathieu, uh, welcome to the show. Welcome me. Thank you.
6
00:00:50.500 --> 00:01:03.900
Mathieu, you had a very impressive, uh, career in data and technology. Uh, can you tell us a bit more what initially spiked you, uh, in data management and what keeps you motivated in this, uh, ever-evolving field?
7
00:01:04.620 --> 00:01:12.580
I'm blushing a bit now. Thank you for the compliment, and yeah, we've, we've got a similar sort of background, so I'll have to return the compliment as well. Thank you. Um,
8
00:01:13.800 --> 00:01:24.740
yes, some of my early projects were actually squarely in the structured data world, so e-commerce and quoting platforms, but, but always was a industry and practical focus.
9
00:01:26.040 --> 00:01:37.440
My first real foray in data management as, as we think about it today, was actually in nineteen ninety-nine when I was the co-founder of a, a startup looking at
10
00:01:38.440 --> 00:01:51.740
data management and data exchange in the manufacturing supply chain. So ever since, I've looked at data from a how do we employ data, how do we leverage data, how do we operationalize data to do things differently.
11
00:01:52.520 --> 00:02:03.760
So a-and I've explored all aspects of it. Search engine, taxonomy and ontology management, classification, metadata management, knowledge graph. Um, so actually from a...
12
00:02:03.860 --> 00:02:14.680
You know, you asked what is my motivation, what, um, what is my vision. Um, it's to help you understand and act on the data you have. Yeah.
13
00:02:14.960 --> 00:02:22.860
I, I, I understand that you're not into the data space because you came from the technical background data, but more from the business side.
14
00:02:22.910 --> 00:02:30.720
So you feel if you have the data, those insights, how they can help, uh, the logistics, uh, supply chain, where you came from.
15
00:02:31.340 --> 00:02:41.780
Uh, that's, that's your different take, what you have on, on the data management, uh, approach? I suppose. I mean, it's, it's about the practicality of it, right?
16
00:02:41.840 --> 00:02:50.420
I mean, yeah, I, I've moved, you know, throughout the years, you know, from research, and I actually started back in the days of BT Labs i-in the research department.
17
00:02:50.480 --> 00:02:59.080
So, you know, I, I, I've seen both sides, and we actually still have both sides, uh, in our teams because, you know, keeping abreast with latest technology, et cetera, is really important.
18
00:02:59.120 --> 00:03:05.460
But I think democratizing those technology and making it useful i-is really critical as well.
19
00:03:05.500 --> 00:03:16.680
You know, sometimes, you know, when I discuss with, you know, prospects or customers or partners a-and they look at the data science element, which is really interesting and fascinating. Don't get me wrong.
20
00:03:16.740 --> 00:03:27.340
I mean, I, I love it as well. Um, but I always need to have that, that vision of how it's going to make a difference for someone and how we can drive it to actually simplicity as well.
21
00:03:27.780 --> 00:03:40.570
You know, and, and, and I know we're probably going to explore gen AI and those sort of things later on in, in our, uh, little discussion here, but yeah, often in those conversation, we go hyper-technical really quickly,
22
00:03:41.580 --> 00:03:54.920
a-and we forget that often actually we're, we're here to solve a business problem, and a business problem is led by a business analyst or, you know, by a SME, by someone who runs their operation, and, and it's not a data specialist, not an IT specialist.
23
00:03:54.980 --> 00:04:01.100
I mean, they want to be able to use data for their process, but they're coming at it from a process perspective. Yeah.
24
00:04:01.180 --> 00:04:08.300
It's, it's nice that you already touch up-upon gen AI and, and, and artificial intelligence, the large language models, and the semantics.
25
00:04:08.920 --> 00:04:20.399
What, what, what is your feeling on the current state of data management with this evolution, uh, with AI and supporting data management, supporting decision, uh, driven organizations?
26
00:04:20.740 --> 00:04:34.360
Uh, and what are the biggest challenges organizations are facing, uh, these days? We're at a very interesting juncture in time. We, we've got two forces. I mean, one is economics,
27
00:04:35.580 --> 00:04:42.400
and, you know, we clearly have seen a, a, a fairly drastic change, uh, in, you know, the economy around us in the last few years.
28
00:04:43.040 --> 00:04:50.740
And when, you know, once upon a time, if you want to make more revenue, you were going out and selling more. Uh, selling more is actually much harder now.
29
00:04:50.800 --> 00:04:55.410
So people are trying to make the most of what they have, you know. And, and, and what does that entail?
30
00:04:55.800 --> 00:05:02.680
Well, that entail analyzing your data, understanding your market, understanding your own behavior, optimizing, you know, better customer intimacy.
31
00:05:02.720 --> 00:05:13.140
So a lot of those things are actually data-driven, very, very data-driven. So moving from a, you know, let's go out and sell to let's analyze our data and operationalize it and make decisions based on our data.
32
00:05:13.180 --> 00:05:23.120
So I'd say, you know, a very interesting inflection point on how organizations are thinking about the data. The second inflection point clearly is technology.
33
00:05:23.140 --> 00:05:31.300
You know, we've already mentioned generative AI twice in five minutes. [laughs] No doubt it'll come up again. Um, but the...
34
00:05:32.080 --> 00:05:46.460
Yeah, it has an associated impact on how do we prepare data, how do we leverage data and bring data in that space. So what are the challenges here? Well, we've got a big need to change culture around data.
35
00:05:46.560 --> 00:05:49.650
Um, and that goes, you know, from building,
36
00:05:50.700 --> 00:06:04.860
really building our data stewards and empowering them to identifying and, you know, giving the right tools to our data citizens, to thinking about architectures differently.Yeah.
37
00:06:05.130 --> 00:06:15.380
But that, that's a lot what is, what is going on. I think you made a smart remark on where back in the days when you wanted to make more revenue, you had to sell more.
38
00:06:15.390 --> 00:06:23.210
Does that mean now with technology and data and the insights, we are capable of leveraging those margins and, and improving that?
39
00:06:23.240 --> 00:06:32.620
So really into the business process optimization part, thanks to the technology and the data, if I understand what you're saying. So that's definitely one aspect.
40
00:06:33.300 --> 00:06:44.220
Um, so, you know, optimizing processes, um, interactions, you know, supply chain, et cetera, all of that can be driven by data. Uh, I think, you know, there are other dimensions as well.
41
00:06:44.729 --> 00:06:46.640
You know, micro-targeting of your, of your customers.
42
00:06:46.780 --> 00:06:59.640
You know, when, when you understand your customer need, when you understand your customer interactions, you can provide the right product, the right service at the right time without going for a full blast, uh, to, to, to everyone.
43
00:06:59.740 --> 00:07:09.600
Um, there are also optimization or productivity enhancement by bringing the right data to the right person at the right time.
44
00:07:10.380 --> 00:07:16.880
You know, if, if you think about, you know, a, a lot of knowledge graph, a lot of data fabric work, uh, that we're doing, th-that...
45
00:07:16.940 --> 00:07:36.160
those types of projects are typically about making sure that within a business process, you've got all the data you need to make the decision at the point you need to make it, instead of having to reach out to different team members, you know, um, maybe create an internal support request for someone to go and look back at a contract or find the right data in a si- in a data silo somewhere.
46
00:07:36.220 --> 00:07:45.420
So there's a lot of optimization there, which is not necessarily cost-cutting or anything. It's, it's about enabling people to do their job better. Yeah.
47
00:07:45.460 --> 00:08:00.500
It's, it's cutting down those silos, and the technology helps us better, well, combining those organizations where before you really had to put out that request and then unify it in, in, in your own way.
48
00:08:01.300 --> 00:08:10.310
Uh, MarkLogic, uh, the company you came from, which has now been acquired by Progress, is a leader in unifying that, uh, that data and integrating that data.
49
00:08:10.760 --> 00:08:24.120
Can you explain the unique approach that, that MarkLogic, uh, took, uh, on integrating that data? And what are the advantages now of being embedded in the, in the Progress, uh, data platform ecosystem? Sure.
50
00:08:25.220 --> 00:08:34.780
I mean, a-as you just mentioned, yeah, MarkLogic and the Progress data platform now, you know, it's like we're building around it. You know, we...
51
00:08:35.229 --> 00:08:47.130
Our objective here is to really unlock the value from complex data, you know, to make it, you know, easily discoverable, fit for purpose, you know, to build high-value, high-power applications,
52
00:08:48.180 --> 00:08:54.300
uh, nowadays to deliver decision intelligence and support your AI journey, of course. Uh, yeah, what do we do behind the scene?
53
00:08:54.340 --> 00:09:00.479
I mean, we integrate data wherever it, it is coming from, so, you know, breaking down those data silos.
54
00:09:00.520 --> 00:09:13.480
We're actually storing the part of the data we care about, so usually metadata, knowledge graph, entity, those things, et cetera. You know, we provide querying and, you know, retrieval capability on top of that.
55
00:09:14.580 --> 00:09:28.160
We have really powerful way to curate, govern, and secure data. And because, you know, storing data, managing data, and reaching data is one thing, uh, but using it is another.
56
00:09:28.600 --> 00:09:41.860
I mean, we have-- we are building and have built over the years, you know, really good capabilities around exploring and analyzing that data. So what are the, the pivot, the, the pillars of our technology?
57
00:09:41.900 --> 00:09:52.320
I mean, first of all, a truly multi-model database. You know, we can manage documents, graph, relational, geospatial, semantic, bitemporal data in a single engine.
58
00:09:52.360 --> 00:10:01.500
You know, we have really high performance search and query capability, so we can run complex queries across all of your data and get really accurate and sub-second results.
59
00:10:02.480 --> 00:10:14.140
We have powerful data integration and data management capability, so integrating data. We have that's, you know, the DataDirect part of the portfolio, which is part of the platform.
60
00:10:14.480 --> 00:10:25.320
Um, really impressive way of curating data and adding semantics and, you know, metadata management, uh, with Semaphore, um, agile modeling, et cetera.
61
00:10:25.460 --> 00:10:32.620
So we have smart mastering in MarkLogic as well as part of the Data Hub, um, and, and Data Hub Central product.
62
00:10:33.200 --> 00:10:45.140
So really powerful metadata and data management capability overall complemented by a comprehensive suite of tools for developers and business users. You know, why?
63
00:10:45.180 --> 00:10:54.160
Because, A, we want developers to build applications faster, um, to reduce the time to market. And why business users?
64
00:10:54.180 --> 00:11:04.440
Because as we mentioned, uh, as I mentioned already several times, you know, we want to put our tools, our capabilities in the hands of the people that run the business with their data.
65
00:11:05.420 --> 00:11:12.140
I mean, it, it, it's really important that a data project, it doesn't go at the pace of an IT project.
66
00:11:12.780 --> 00:11:22.120
We can power IT for everything to do with the integration, running the platform, but we need the data to be in the hands of the business users to move at the pace of the business.
67
00:11:24.500 --> 00:11:35.060
Are you referring that there is a different speed between IT and the business part? I mean, never heard about that, [laughs] but... No, no, I'm just, just laughing, but I do understand what, what you mean.
68
00:11:35.160 --> 00:11:47.600
So you provide as well with, with the Progress platform a way of businesses to easily, uh, create their, uh, data products and make data more accessible. Then you put it through IT.
69
00:11:47.740 --> 00:11:58.600
I mean, IT has its value by governing it, uh, by providing the architecture, making sure it runs twenty-four/seven, it's stable. I mean, that needs to be done.
70
00:11:58.790 --> 00:12:05.640
Uh, sometimes if you see it comes from the business, it's not so solid as it should be. It's not so secure as it should be.
71
00:12:05.740 --> 00:12:13.026
So that's the part where I see IT still needs to be-Knocking on the door and say, "Hey, guys, we need to slow down."
72
00:12:13.216 --> 00:12:22.696
Slowing down, not because of we do things so slow, but we need to do these, these things to make solid solutions in such a way. A-absolutely. Absolutely.
73
00:12:22.736 --> 00:12:32.196
I mean, the, the infrastructure part, you know, all the plumbing, the connections, the integration, all of those things, when those have to be governed, you know, go through a, a normal IT life cycle, et cetera, et cetera.
74
00:12:33.596 --> 00:12:46.776
Where many data projects break down today is when the data and the data interpretation is actually treated as a data- as a, you know, project life cycle artifact. I mean, data moves at a much faster pace.
75
00:12:46.826 --> 00:12:55.326
Its interpretation is different. We need... Essentially, we're bringing data agility. Kinda in, in the same way that, you know, engineering teams have been moving to an agile model,
76
00:12:56.916 --> 00:13:10.756
so you can deliver incremental functionality, you know, shippable incremental functionality at, at the end of this, of each sprint, and you're supported by a shared IT organization that makes sure infrastructure on which you run is, you know, secure, governed, et cetera.
77
00:13:11.596 --> 00:13:14.956
We want to do the same thing with data, all right? Build data agility.
78
00:13:15.596 --> 00:13:30.276
In-instead of engineering scrum team, you've got business scrum team, thinking about a data product, thinking about how they need to interpret the data, use the data in their operation, supported by an IT organization which provide them with the software, you know, and the infrastructure to operate in.
79
00:13:31.396 --> 00:13:57.776
Does that mean that the data platform from Progress and all the toolings around that now provide the two different modi where you can have the exploration part, which goes pretty fast on the speed of data, and that you can, once you have that data product developed or that vision, uh, for the data or the decision you wanna take with your data, and then move it through the, the slower path from IT, be fully governed, uh, fully secure, and you can easily switch between the two, uh, of those?
80
00:13:58.616 --> 00:14:16.736
I, I'd go even further than that because we completely separate the infrastructure part of the software, which is, you know, governed, et cetera, and goes through the IT life cycle as normal, from the data part, which is using a more information science, data science librarian-type, uh, workflow.
81
00:14:17.316 --> 00:14:29.076
So without changing the software configuration, you can change the modeling and an ontology and try a new data scenario, try a new data interpretation in a matter of hours, in a matter of minutes, in fact.
82
00:14:29.776 --> 00:14:47.396
Um, and, and not only that, but, you know, from a technology perspective, because it is all semantics and W3C standard compliant, you can actually have several of those interpretation running side by side, and even at query time, you can say, "Oh, I want to run through the lens of this world or that business domain or that process."
83
00:14:47.916 --> 00:14:59.846
So there's no, you know, hard-coded schemas. There's no hard-coded, you know, view of the world. There's no, you know, predefined, you know, data reduction. And I think data- Yeah. It's-...
84
00:14:59.856 --> 00:15:04.976
data reduction, loss of precision is actually something that, um, you know, we talk a lot about, um, in our project.
85
00:15:05.045 --> 00:15:15.446
'Cause earlier in the cycle, you do data interpretation, typically, the more, [sighs] the more information you lose. Yeah. It's, it's data agility.
86
00:15:15.856 --> 00:15:26.056
And when you say we, we can through-- look through the different lenses, we, we-- Well, I'm the traditional schooled data warehouse builder, dimensional modeling, uh, data guy.
87
00:15:26.636 --> 00:15:36.356
Uh, and we've b- had the discussion of one version of the truth where Bill Inmon, one of my former guests as well, uh, was very proud of, of providing that.
88
00:15:36.646 --> 00:15:50.846
But we've learned over time that there are different versions of the truth, depending on which department or where you are in, in your environment, and, and that's important that you can easily switch those views, but still on, on the same basis, the same foundation.
89
00:15:51.756 --> 00:16:02.896
Exactly. Because based on who you are, where you are, you might have varied different information needs, even though your query is the same. It might be based- Yeah...
90
00:16:02.906 --> 00:16:12.526
on access, it might be based on local translations, it might be based on, you know, regulations, all sorts of things. I mean, we, we do a lot of projects in the pharmaceutical space.
91
00:16:13.076 --> 00:16:24.236
Well, even if the research data is the same, even if the molecule, uh, is supposedly the same, it might not be manufactured in the same way because the recipe might change depending on the local ingredients.
92
00:16:24.246 --> 00:16:32.866
The brand might also be different depending on the market. The reporting requirements may be different. The little leaflets you get in the box may be different because that's based on local regulations.
93
00:16:33.326 --> 00:16:36.276
It doesn't mean the product is different. I mean, conceptually, it's treating the same thing.
94
00:16:37.156 --> 00:16:48.876
Um, people might not have the same symptom because symptoms are also something which varies depending on, you know, where in the world you live. So it's interesting. I mean, we, we could go on for, for hours on that.
95
00:16:48.976 --> 00:16:59.936
But if you, if you think, if you go back all the way to the root of ontologies and, you know, semiotic, not semantic, but semiotic, the relationship between the words we use, the objects in the real world,
96
00:17:00.896 --> 00:17:13.996
and the, the concept behind them, you know, th-that mapping varies drastically depending on audience. Yeah. F-fully understand what, what, what, what you're, uh, talking about.
97
00:17:14.316 --> 00:17:20.136
Now, with, with generative AI and the large language models coming in, and we're talking about different perspectives.
98
00:17:20.196 --> 00:17:35.616
I mean, people have used ChatGPT already for the last year, and where you s-- Well, there it's pretty clear that you can add a different role, and depending on the role you give it, uh, to the large language model, you get a different outcome.
99
00:17:35.866 --> 00:17:44.556
But telling you the same thing, but depending on, on the role you give it, you get a different wording. That's, that's exactly what, what you're referring to, I think.
100
00:17:45.376 --> 00:17:59.136
And how do you see these large language models will, will help, uh, revolutionizing, um, the, the space of data management? Will that help us in, in building better solutions, uh, more valuable solutions?
101
00:18:00.116 --> 00:18:04.436
I mean, it is profoundly changing the field of data management. I mean, no doubt about it.
102
00:18:04.776 --> 00:18:13.456
I mean, from a technical perspective, because it gives you new tools and new approaches, um, you know, new ways of interpreting a, a corpus, new way of serving a corpus.
103
00:18:14.596 --> 00:18:26.932
Um-From my perspective, it's also really interesting because it is driving a very different type of adoption for metadata and data management. I mean,
104
00:18:28.232 --> 00:18:38.342
a lot of customers, a lot of prospects are coming to us saying, "Hey, we've started this generative AI project, but it's not really performing the way we want." Why?
105
00:18:38.632 --> 00:18:46.622
Well, what have you given it as an input, you know, for, for RAG strategy, et cetera? It's like, well, the data we have. Well, guess where the problem lies, right?
106
00:18:46.792 --> 00:18:55.332
So, you know, from an adoption perspective, generative AI is really interesting because, you know, it is fueling the need for better data.
107
00:18:57.292 --> 00:19:08.332
Our customers, our prospects, our partners are all telling us, you know, of horror story where customers have tried to use their data as is to drive a generative AI solution. You know, same old, same old.
108
00:19:08.792 --> 00:19:16.512
You know, I'm not going to say garbage in, garbage out because actually, the generative AI solutions are making a good job at hiding some of that.
109
00:19:16.532 --> 00:19:25.812
But the reality is, if you want precise and accurate and trustworthy answers on private data, today, you need a RAG strategy.
110
00:19:25.852 --> 00:19:34.892
Today, you need, you know, the right level of security, the right level of understanding, the right level of search. You know, and that goes all the way back to our story of breaking the data silos.
111
00:19:35.712 --> 00:19:46.052
You know, moving from a application-centric architecture to data-centric architecture, having, you know, knowledge graphs, having a data fabric from which you can actually surface the right data at the right time.
112
00:19:46.652 --> 00:19:52.742
You know, and the conversation we just had about micro-targeting our users, about understanding our users' behavior,
113
00:19:53.792 --> 00:20:09.032
I mean, all of that goes right back in the way we search our fabric to include the right fragment in the RAG, you know, to support the generative AI response generation. So from that adoption perspective, we're seeing a
114
00:20:09.952 --> 00:20:15.372
renewed interest in data management and in metadata management and, and search technologies in general.
115
00:20:16.752 --> 00:20:24.352
The-- Associated with that, I suppose, you know, there's a more social or societal perspective, which is data is becoming cool again.
116
00:20:25.032 --> 00:20:33.532
You know, you, you hear of people, you know, trying things with ChatGPT, trying their own, you know, AI, ML pipelines and MLOps, and it, it's great actually.
117
00:20:33.572 --> 00:20:43.112
I mean, you know, socially, I'm now answering questions about, you know, stuff we do in our research labs, and it's actually really interesting. People didn't pay attention ten years ago.
118
00:20:43.552 --> 00:20:59.302
[laughs] However, uh, you know, you were asking about impact and changes. One big impact and change we're seeing actually in the way we interact, um, on a lot of data projects is a, a
119
00:21:00.492 --> 00:21:14.572
realization of the legal and threats associated with how data is used. So, you know, DLP, uh, privacy, all of those things are fueling actually a lot of additional inquiries.
120
00:21:15.212 --> 00:21:25.772
Um, you know, a lot of additional time spent on projects designing, uh, what should have been, you know, catered for all along. Uh, but certainly, um, organizations are now understanding
121
00:21:27.632 --> 00:21:34.642
that their data, the minute it's, you know, out in the LM may or may not belong to them, and they may or may not be in control, and there are unknowns.
122
00:21:34.832 --> 00:21:45.052
So we're seeing a lot of activity in, um, you know, sec- with security teams, with legal teams, with procurement teams around, you know, how do you handle, um, this new landscape?
123
00:21:47.172 --> 00:22:00.092
So what you're seeing here is that previously we had to pay attention to the governance and the security of the data, but because it was more siloed, it went lesser out, let's call it like that.
124
00:22:00.792 --> 00:22:09.692
Uh, people were not paying so much attention to it. I recall the days where you said, "Okay, I need access to the server, to the database. Here are the admin rights, full open, whatever you do with it."
125
00:22:10.612 --> 00:22:21.871
Uh, you could do as a developer kind of everything on, on the server, but you're very well aware of the security breaches of the privacy regulations and all of that.
126
00:22:21.932 --> 00:22:34.311
What you need to take into account, uh, and be careful with because you see so much, uh, valuable information, uh, which shouldn't be for all eyes, but there is a kind of trust towards developers to develop that product.
127
00:22:34.792 --> 00:22:50.272
But now with the large language models, this information is easier to go out of the organization, so people become more aware of, of governing that, of protecting, uh, the privacy of, of all the information which goes into the model.
128
00:22:50.492 --> 00:22:58.472
Uh, and that's where they now become more aware and paying more attention to, to that part of, of data management. Absolutely.
129
00:22:58.512 --> 00:23:09.672
So, you know, until relatively recently, the majority of security work associated with project was, you know, doing role-based control, compartment security, et cetera, within the database.
130
00:23:10.152 --> 00:23:18.882
Making sure that the data in the system or in the s- in, in the fabric, in the search interfaces, you know, we're reaching the right person.
131
00:23:19.592 --> 00:23:31.312
You know, now we're seeing a lot of organization go beyond that and either contractually or through systems, making sure that once that data is served in the application, there's actually traceability beyond that.
132
00:23:31.372 --> 00:23:36.972
If it's sent out to a customer, uh, if it's actually sh-shared with another coworker, you know, all of those sort of things.
133
00:23:37.052 --> 00:23:49.852
Um, and, and especially, you know, restrictions around using that data in training LLMs, in AI, um, or machine learning projects, et cetera. So, so you see,
134
00:23:50.832 --> 00:24:04.312
there's definitely a recognition now that using data for the purpose of training has risks. It doesn't mean we shouldn't do it. It means that we should understand the full scope
135
00:24:05.552 --> 00:24:19.652
and accept the risk when we want to accept the risk. So it's a different approach. It's-- I, I feel like when you say we, we want to... You have two s- two modi again in, in developing solutions.
136
00:24:19.732 --> 00:24:29.380
If you want to be in-innovative, creativeYou have to-- Well, you're building less secure and governance solutions because you wanna go fast, and that is slowing you down.
137
00:24:29.880 --> 00:24:41.260
And on the other hand, if you want to have it protected, that takes ages before you have the regulations in place and, and so on. So that's, that's a bit what I see where you have to find a trade-off.
138
00:24:41.600 --> 00:25:03.060
Do we use that data to take the risk to go faster to build really an innovative product as such, or do we really put the foot on the brake and, uh, within the next one hundred years, maybe we have the next [laughs] large language model, and not at the speed where you see every week you have a new version of a, of an open source large language model.
139
00:25:03.480 --> 00:25:15.900
Technology goes at such a fast pace. I mean, I've been diving into large language models now in-depth into for the last six months, and if you see, you take frameworks like, like LangChain, it changes every week.
140
00:25:15.960 --> 00:25:21.940
Documentation changes, fra- uh, the complete framework changes after a month, and then you kinda...
141
00:25:22.020 --> 00:25:28.500
Well, it's, it's hard to stay on top of it on the development and on everything what is on the market to build those solutions.
142
00:25:28.510 --> 00:25:42.140
And I think that's, that's gonna be a challenge now for, for most of the people to understand, one, on what is possible with the technology, but as well on capturing that in, that, that capability and using that to build the innovative products.
143
00:25:43.700 --> 00:25:57.840
So talking about that, how do you see the, the data management space evolving over the, the next to five, uh, five to ten years, and, and what emerging trends, uh, should data professionals and executives, uh, should be paying attention to?
144
00:26:00.900 --> 00:26:15.300
Well, data quality, of course, is, uh, a big one for me. Data management, right? [laughs] Metadata management as well. Um, so five to ten years, I mean, [sighs] you've seen the, the pace of change right now.
145
00:26:15.420 --> 00:26:20.160
I mean, uh, uh, in some of those topics, it's hard to predict what's going to come in two weeks. So, um,
146
00:26:21.080 --> 00:26:35.900
this said, and, and back to what we said a, you know, a few minutes ago, there's this very big dichotomy, you know, difference in, in swim lane, in pace between our industry and manufacturing and, you know, et cetera users, regardless of the, of the industries they're in,
147
00:26:36.880 --> 00:26:45.760
and our research. Uh, so if you look at projects in production or, you know, projects that are dealing with data at scale, I think we're still,
148
00:26:46.980 --> 00:26:50.800
you know, in the early days of data fabric and knowledge graph and, a-and all of those things.
149
00:26:50.820 --> 00:27:01.780
So there's still, you know, a lot of technology to be developed, a lot of technology, a lot of processes to be simplified, uh, to make it applicable and make sure it's rolled out and successful into the environment.
150
00:27:01.820 --> 00:27:13.400
I mean, we're still seeing the same three big hurdles, um, you know, for, for our customers, which is the plumbing, uh, the data quality, and actually operationalizing your data.
151
00:27:13.469 --> 00:27:21.860
So, you know, that's the, the order in which as an organization you need to, to, to jump, um, and, and the order in which you're, you're falling as well.
152
00:27:21.920 --> 00:27:38.200
And, and just something like plumbing, I mean, you just said it a minute ago. LLMs, new models coming out on a weekly basis, right? The modern data stack is fantastic. You know, it's great, it's composable.
153
00:27:38.240 --> 00:27:47.940
But for many, it is too composable, and they don't know where to go anymore. Um, you know, and, and that results in projects which are... You know, it's fantastic for research projects.
154
00:27:48.240 --> 00:27:56.940
For, for production projects, you know, what framework do I use? What technology? What API? I mean, it's changing on a weekly basis, on a daily basis in some cases. Hold on, I've, I've just come in this morning.
155
00:27:56.980 --> 00:28:08.240
My API, you know, my call doesn't work anymore. What, what's going on? So, you know, one thing that we're really working on, um, in, in the Progress data platform is making sure we've got an end-to-end story.
156
00:28:08.260 --> 00:28:10.680
Yeah, so you don't have to reinvent that plumbing.
157
00:28:10.740 --> 00:28:28.120
And, you know, if you think about the strengths of Semaphore, the strengths of MarkLogic, the strengths of DataDirect, the strengths of Corticon, you know, the strengths of Kendo, another Progress product that we're using for visualization, you know, we have all of the elements to really make high power, you know, high value applications.
158
00:28:29.500 --> 00:28:38.060
And the data platform strategy is essentially creating the blueprint. So as a customer, you don't have to reinvent the wheel. You can follow a proven recipe.
159
00:28:38.980 --> 00:28:45.440
So your data projects can be integrating with the data sources all the way to delivering value in a business application.
160
00:28:46.980 --> 00:28:55.620
Yeah, as f- as soon as possible, as fast as possible, minimum time to market, minimum complexity, you know, minimum investment in the commodity, in the plumbing.
161
00:28:56.300 --> 00:29:05.890
Because where the value lies from a business perspective is in the data quality. You understand your data. You are producing that data for a reason. So what is that reason? Can you model it?
162
00:29:06.640 --> 00:29:19.040
Can you actually identify it in content? And how can you leverage it to change the way you do business? You know, and, and very often, projects are, you know, [sighs]
163
00:29:20.940 --> 00:29:32.240
completely submerged in the quagmire of choosing the right moving API, you know, uh, or, and the right component for the day. Then they completely lose track of, of the rest of the data journey.
164
00:29:33.960 --> 00:29:44.880
Now, of course, I imagine there'll be some stabilization, some consolidation, and, you know, we're already starting to see that consolidation. Um, but we need to provide, I think, uh, as a,
165
00:29:46.500 --> 00:29:59.000
you know, as an industry, actually, some guidance on how to compose the right architecture to achieve the right data quality, so you can operationalize your data. All the while, maintaining security
166
00:30:00.800 --> 00:30:06.600
and helping you enforce governance. At the end of the day, without governance, you won't have trustworthy data.
167
00:30:07.560 --> 00:30:19.700
Without security, you have all sorts of breaches, repu-reputational issue, and you'll be caught up, you know, caught up in the crossfire of solving, you know, data security breaches rather than solving your business problems, right?
168
00:30:19.800 --> 00:30:34.442
So back to square one. [laughs]We, we have a lot of more possibilities because we have so much choice of technology, but it made it only more complex. Uh, so that's, that's my takeaway from what you just said.
169
00:30:35.112 --> 00:30:45.512
But as a, as a vendor, you're helping customers as well with the blueprints where you say, "This is what we've tried. We've put a lot of research in it. We know this solution works."
170
00:30:45.692 --> 00:30:53.132
Maybe you have so many other different vector databases, what is possible. We, we know for this type of problem, this is the best to use.
171
00:30:53.382 --> 00:30:58.152
This is the plumbing you can use, this is the visualization, and this is how you can build your product.
172
00:30:58.672 --> 00:31:11.012
If we see you have a different need, we can revisit and give you a different blueprint and then, uh, migrate from the, from the previous one to this newer one with your new, uh, requirements, uh, if I understand.
173
00:31:11.612 --> 00:31:22.592
Ab-absolutely. And one thing I'll, I'll probably jump on. You just said, "This is the best one. This is the best one. This is the best one." Well, actually, you don't necessarily need the best one.
174
00:31:23.472 --> 00:31:31.412
You know, sometimes the best one is actually too complicated, too expensive, too whatever, too hard to manage. I-it's the right one for the use case.
175
00:31:32.372 --> 00:31:59.132
You know, f-for example, we may not have the best vector database ever in MarkLogic, but because we vector as well as, yeah, all of the other data types that we support, all the other query types that we support, all of the other indexing strategy that we support, it's actually the best for the use case because it allows you not only to operationalize your data, do your, you know, analytics, but also serve your data through, um, you know, RAG for GenAI solution.
176
00:31:59.192 --> 00:32:12.492
So it's about how do we optimize the architecture and the product capabilities and make the right choices on the way to deliver an enterprise solution. Yeah. I mean, it's kind of flashback.
177
00:32:12.552 --> 00:32:24.372
I mean, I've just built a large language model, uh, solution as well, and I had so many choices, like you said, the vector databases, the transformers, the embeddings, uh, the all of that part.
178
00:32:24.452 --> 00:32:35.972
And because it was very new, I thought, okay, with what I read, I go for just a SQLite database, which is easy accessible. You can look at your data. You can see the metadata, how it works. Come on.
179
00:32:36.032 --> 00:32:40.752
If you didn't use MarkLogic? [laughs] It's too expensive. No. [laughs] Shocking.
180
00:32:40.872 --> 00:32:48.852
Uh, I mean, I think, I think it's a, it's a good use case, definitely, uh, to, to look at MarkLogic with the semantics and, and, and put that in place.
181
00:32:48.972 --> 00:33:03.412
But you're absolutely right where I said, okay, this is what I feel is the best one, and I know, depending on once I have the full solution built, I can revisit all the individual components, and I know where we can optimize if the outcome is not what I'm expecting.
182
00:33:03.472 --> 00:33:11.072
But first I need to build the solution with the individual components and then, uh, value, uh, evaluate that, that part.
183
00:33:11.792 --> 00:33:24.372
So looking at that on, on a more personal note, what values guide you in your personal journey and, and, and how does these influence your vision for the future, uh, for data and analytics?
184
00:33:26.372 --> 00:33:38.952
Well, I think I, I opened with that already. You know, it's making data a-and the knowledge within that data accessible so you can act in informed ways. I mean, we could call that democratizing data management.
185
00:33:40.481 --> 00:33:44.182
Um, you know, it's leveraging your data. That's, you know,
186
00:33:45.392 --> 00:33:55.292
what I've actually tried to achieve and what I've actually worked on in pretty much every single one of my roles since, you know, for the last twenty-five years, essentially.
187
00:33:55.612 --> 00:34:07.472
[laughs] And how does that influence, you know, my, my vision of the future? Well, clearly one-sided and fairly focused on the, on the data management and analytics, but, um, you know, it's about use case.
188
00:34:08.252 --> 00:34:23.592
It-it's about identifying high-value use cases, identifying where data can help. Um, yeah, because if, if you look at the way we operate today, eighty percent of data is unstructured.
189
00:34:25.512 --> 00:34:34.562
The vast majority of data that organizations are producing, exchanging is not used for decision-making.
190
00:34:35.132 --> 00:34:47.232
The, the twenty percent that lives in a relational world, that's what we put on BI dashboards, that's what, you know, our management teams are looking at on their weekly stats and, and scorecards. It's so...
191
00:34:47.292 --> 00:34:57.591
such a small fragment of the overall data landscape we've got to actually make decisions. So, you know, changing that status quo i-is really the motivator here.
192
00:34:57.632 --> 00:35:11.512
You know, and can we bring the full context at the point of decision? And the full context comes from understanding, you know, the full scope of the data and interpreting it, it in the right way.
193
00:35:11.552 --> 00:35:20.442
So, you know, going back to semantics, that allows you to understand the context of the data from, you know, the point of origin. You know, why was it created in, in that way?
194
00:35:21.312 --> 00:35:28.102
And, and some of it is technical, absolutely, and that's why we've got, you know, all the, all the product suites that we have to help you in that data journey.
195
00:35:28.572 --> 00:35:34.452
But a lot of it is also, you know, data science, and a lot of, of it is also library science.
196
00:35:34.512 --> 00:35:44.652
I mean, good old traditional looking at data, you know, talking to the data producers, understanding what they mean, what they, what they've tried to do, why they did it.
197
00:35:45.592 --> 00:35:58.752
And, and sometimes, you know, even today, you-you've got some-sometimes absolutely fantastic answers. It's a, it's a critical mindset. It's, it's that challenging and, and just what you see, not always believing.
198
00:35:59.012 --> 00:36:09.932
I, I see that as well with, we, we have that with the ChatGPT, is you, you put, you, you put a question forward, you get an answer, but you always need to challenge that answer. Is it really right?
199
00:36:10.452 --> 00:36:21.582
And you need to be sure that it is factual correct in, in such a way. So that, that demands a, a critical mindset of, of how you look at your data and, and how you take that information.
200
00:36:21.702 --> 00:36:33.652
It's like the journalists back in the days, they, they were forced to just double, triple verify all the information they got. But now with the speed or at which they need to workIt's, it's getting upended.
201
00:36:33.752 --> 00:36:40.202
Sometimes they get a fact and they think it's true, and give that out, and then don't do the, the double fact-checking.
202
00:36:40.672 --> 00:36:54.112
Now thinking about that, what advice would you give to newcomers that are entering into the data and analytics space and, and make that really their business, their, their roadmap to, to be successful? Be curious.
203
00:36:56.172 --> 00:37:04.102
Yeah, curiosity is actually the number one skill here. You know, because, you know, we often talk about data mining,
204
00:37:05.152 --> 00:37:14.242
you know, as, you know, one of the starting points of creating your data journey, 'cause you need to understand what you have and what it means and, and all those things. You know, you need to be curious about
205
00:37:15.512 --> 00:37:23.092
how data came to be. You know, why was it produced? Why was it produced the way it is? You know, what was meant by the author? What is needed by the audience?
206
00:37:23.472 --> 00:37:28.001
You know, some of it is, you know, business acumen, some of it is technical, some of it is...
207
00:37:28.262 --> 00:37:38.452
You know, the bottom line is if I look at all of my teams, the successful teams are those that are not losing the curiosity over time. You know, keep trying to discover. You know, it's a--
208
00:37:39.632 --> 00:37:52.332
Ultimately, you know, data management is detective work, and you need to be on your feet, you know, looking, scrutinizing all the time. Yeah, everyone could be a data citizen. Everyone is a data citizen.
209
00:37:52.912 --> 00:38:04.332
You know, some people might not know it yet. It's only a matter of time. [laughs] But do need-- do organizations need to provide a bit more time for people to allow that exploration?
210
00:38:04.592 --> 00:38:14.272
Uh, like you say, you say, "Be curious," which is very general, but then you add it onto it, it can be technical, it can be business, it can be data-related.
211
00:38:14.312 --> 00:38:30.252
So it's not only in the domain where you are or are a specialist, but be more holistic in, uh, in being curious across the business where you're working in, uh, in such a way. Absolutely. And, you know, some of the more,
212
00:38:31.952 --> 00:38:42.192
I was gonna say interesting, some of the more successful, uh, projects that we've seen, uh, were completely, you know, organized in ways that are non-traditional.
213
00:38:42.752 --> 00:39:02.312
You know, we, we had a, a really interesting project which started as a almost one-man show, uh, from a, a doctor in a line of business, uh, that was actually a bit puzzled by some of the data that he was observing, and he started building models, you know, doing things in R and doing, you know, doing all sorts of things.
214
00:39:02.352 --> 00:39:10.032
And we, we met him, you know, in, in one of the interview at that client. The client didn't even know that they had such a, a doctor in the organization that was looking at data.
215
00:39:10.052 --> 00:39:15.372
We had some fantastic conversation, and he became one of the data stewards for, you know, one of the, one of the big project.
216
00:39:15.592 --> 00:39:35.532
Brought his experience of the field, brought his experience of, you know, what things mean in the context and how you com- you combine what the data says with the clinical science, with your interpretation of how you interact with a patient to actually do essentially data science, but with the human in the loop and an understanding of the human side of it.
217
00:39:36.212 --> 00:39:44.632
You know, a-and I think that's what I mean by curiosity. Always keep all of the variables, all, all of the dimensions of the project, you know, in mind
218
00:39:45.652 --> 00:39:49.982
when you're, you know, when you're designing solutions, when you're thinking about the results, when you're thinking about the algorithms.
219
00:39:50.032 --> 00:39:59.591
I mean, if you go at it from a purely technical angle, you might get something interesting, but is it true? Is it relevant? Is it going to make a difference? Yeah.
220
00:40:00.182 --> 00:40:13.512
Stay interested, stay motivated, and stay energized and, and, and you will find your, uh, answers of, of what you're looking for. Uh, Mathieu, we-we're coming to the end of, of the show. Uh, I always close up with
221
00:40:15.052 --> 00:40:28.091
data is connecting us all, but music is connecting us all as well, so I would love to hear what is your favorite band or favorite type of music? It's a, it's a very hard question, 'cause I've got quite an eclectic taste.
222
00:40:28.152 --> 00:40:38.162
So I'll, I'll give you a f- a few elements of answer. So I'm actually a, a saxophone player. I don't play a lot. Uh, whenever I got time, I, I, I try to do a, a little bit still.
223
00:40:38.792 --> 00:40:54.732
Uh, right now I've got couple of scores on my, uh, you know, on my stand, uh, one from Paolo Conte and one from, uh, Astor Piazz- uh, Piazzolla, so Cafe 1930. Uh, really interesting. Love them.
224
00:40:55.012 --> 00:41:03.352
Uh, this said, you know, on the other end of the scale, uh, tonight is actually the Brave Festival, and we're going with the family to, to watch Sting. So [laughs] there you go.
225
00:41:03.372 --> 00:41:15.072
[laughs] Oh, he's still performing, so that's interesting to h- to hear. Yeah, yeah, yeah. That's, that's quite a very, uh, va- uh, variation on taste of music.
226
00:41:15.132 --> 00:41:23.932
But, uh, really, uh, ex-- well, discovered a lot of, uh, music with my guests which I've never heard before, so this is the fun part for me to close out.
227
00:41:24.312 --> 00:41:37.752
Uh, Mathieu, it was so, uh, much fun having you and interesting, uh, on the Data Gurus, uh, Strategy Podcast. So thank you for coming and having this, uh, very nice chat. Thank you for having me. Always a pleasure.
228
00:41:37.812 --> 00:41:48.032
Thank you for joining us on this awesome podcast. As senior executives, data and analytics architects, and AI professionals, your time is valuable and we appreciate you choosing to hang out with us.
229
00:41:48.072 --> 00:41:54.552
If you liked what you heard, please give us a thumbs up, hit the subscribe button, and leave a comment. We love hearing from our audience.
230
00:41:54.592 --> 00:42:10.091
Don't forget to spread the word on social media, and let's continue to drive innovation in the industry together. Thanks for listening, and we'll catch you on the next episode. [outro music]
1
00:00:00.120 --> 00:00:10.080
Welcome to the Data Strategy Gurus podcast. In this show, we bring together the brightest minds in the world of data strategy, data management, artificial intelligence, and disruptive technologies.
2
00:00:10.320 --> 00:00:16.860
Thought leaders and experts share their insights, knowledge, and experience on how to stay ahead of the game in an ever-evolving data landscape.
3
00:00:17.160 --> 00:00:23.540
Whether you're a data professional, a business leader, or simply someone who is passionate about the power of data, this podcast is for you.
4
00:00:23.820 --> 00:00:35.180
So sit back, relax, and join us on a journey to explore the world of data, analytics, artificial intelligence, tech, and beyond. Hi, and welcome to the Data Strategy Gurus podcast.
5
00:00:35.580 --> 00:00:49.340
F- today we have the VP of technology, Mathieu Jamblé, a former MarkLogic, but now integrated with the Progress data platform. Mathieu, uh, welcome to the show. Welcome me. Thank you.
6
00:00:50.500 --> 00:01:03.900
Mathieu, you had a very impressive, uh, career in data and technology. Uh, can you tell us a bit more what initially spiked you, uh, in data management and what keeps you motivated in this, uh, ever-evolving field?
7
00:01:04.620 --> 00:01:12.580
I'm blushing a bit now. Thank you for the compliment, and yeah, we've, we've got a similar sort of background, so I'll have to return the compliment as well. Thank you. Um,
8
00:01:13.800 --> 00:01:24.740
yes, some of my early projects were actually squarely in the structured data world, so e-commerce and quoting platforms, but, but always was a industry and practical focus.
9
00:01:26.040 --> 00:01:37.440
My first real foray in data management as, as we think about it today, was actually in nineteen ninety-nine when I was the co-founder of a, a startup looking at
10
00:01:38.440 --> 00:01:51.740
data management and data exchange in the manufacturing supply chain. So ever since, I've looked at data from a how do we employ data, how do we leverage data, how do we operationalize data to do things differently.
11
00:01:52.520 --> 00:02:03.760
So a-and I've explored all aspects of it. Search engine, taxonomy and ontology management, classification, metadata management, knowledge graph. Um, so actually from a...
12
00:02:03.860 --> 00:02:14.680
You know, you asked what is my motivation, what, um, what is my vision. Um, it's to help you understand and act on the data you have. Yeah.
13
00:02:14.960 --> 00:02:22.860
I, I, I understand that you're not into the data space because you came from the technical background data, but more from the business side.
14
00:02:22.910 --> 00:02:30.720
So you feel if you have the data, those insights, how they can help, uh, the logistics, uh, supply chain, where you came from.
15
00:02:31.340 --> 00:02:41.780
Uh, that's, that's your different take, what you have on, on the data management, uh, approach? I suppose. I mean, it's, it's about the practicality of it, right?
16
00:02:41.840 --> 00:02:50.420
I mean, yeah, I, I've moved, you know, throughout the years, you know, from research, and I actually started back in the days of BT Labs i-in the research department.
17
00:02:50.480 --> 00:02:59.080
So, you know, I, I, I've seen both sides, and we actually still have both sides, uh, in our teams because, you know, keeping abreast with latest technology, et cetera, is really important.
18
00:02:59.120 --> 00:03:05.460
But I think democratizing those technology and making it useful i-is really critical as well.
19
00:03:05.500 --> 00:03:16.680
You know, sometimes, you know, when I discuss with, you know, prospects or customers or partners a-and they look at the data science element, which is really interesting and fascinating. Don't get me wrong.
20
00:03:16.740 --> 00:03:27.340
I mean, I, I love it as well. Um, but I always need to have that, that vision of how it's going to make a difference for someone and how we can drive it to actually simplicity as well.
21
00:03:27.780 --> 00:03:40.570
You know, and, and, and I know we're probably going to explore gen AI and those sort of things later on in, in our, uh, little discussion here, but yeah, often in those conversation, we go hyper-technical really quickly,
22
00:03:41.580 --> 00:03:54.920
a-and we forget that often actually we're, we're here to solve a business problem, and a business problem is led by a business analyst or, you know, by a SME, by someone who runs their operation, and, and it's not a data specialist, not an IT specialist.
23
00:03:54.980 --> 00:04:01.100
I mean, they want to be able to use data for their process, but they're coming at it from a process perspective. Yeah.
24
00:04:01.180 --> 00:04:08.300
It's, it's nice that you already touch up-upon gen AI and, and, and artificial intelligence, the large language models, and the semantics.
25
00:04:08.920 --> 00:04:20.399
What, what, what is your feeling on the current state of data management with this evolution, uh, with AI and supporting data management, supporting decision, uh, driven organizations?
26
00:04:20.740 --> 00:04:34.360
Uh, and what are the biggest challenges organizations are facing, uh, these days? We're at a very interesting juncture in time. We, we've got two forces. I mean, one is economics,
27
00:04:35.580 --> 00:04:42.400
and, you know, we clearly have seen a, a, a fairly drastic change, uh, in, you know, the economy around us in the last few years.
28
00:04:43.040 --> 00:04:50.740
And when, you know, once upon a time, if you want to make more revenue, you were going out and selling more. Uh, selling more is actually much harder now.
29
00:04:50.800 --> 00:04:55.410
So people are trying to make the most of what they have, you know. And, and, and what does that entail?
30
00:04:55.800 --> 00:05:02.680
Well, that entail analyzing your data, understanding your market, understanding your own behavior, optimizing, you know, better customer intimacy.
31
00:05:02.720 --> 00:05:13.140
So a lot of those things are actually data-driven, very, very data-driven. So moving from a, you know, let's go out and sell to let's analyze our data and operationalize it and make decisions based on our data.
32
00:05:13.180 --> 00:05:23.120
So I'd say, you know, a very interesting inflection point on how organizations are thinking about the data. The second inflection point clearly is technology.
33
00:05:23.140 --> 00:05:31.300
You know, we've already mentioned generative AI twice in five minutes. [laughs] No doubt it'll come up again. Um, but the...
34
00:05:32.080 --> 00:05:46.460
Yeah, it has an associated impact on how do we prepare data, how do we leverage data and bring data in that space. So what are the challenges here? Well, we've got a big need to change culture around data.
35
00:05:46.560 --> 00:05:49.650
Um, and that goes, you know, from building,
36
00:05:50.700 --> 00:06:04.860
really building our data stewards and empowering them to identifying and, you know, giving the right tools to our data citizens, to thinking about architectures differently.Yeah.
37
00:06:05.130 --> 00:06:15.380
But that, that's a lot what is, what is going on. I think you made a smart remark on where back in the days when you wanted to make more revenue, you had to sell more.
38
00:06:15.390 --> 00:06:23.210
Does that mean now with technology and data and the insights, we are capable of leveraging those margins and, and improving that?
39
00:06:23.240 --> 00:06:32.620
So really into the business process optimization part, thanks to the technology and the data, if I understand what you're saying. So that's definitely one aspect.
40
00:06:33.300 --> 00:06:44.220
Um, so, you know, optimizing processes, um, interactions, you know, supply chain, et cetera, all of that can be driven by data. Uh, I think, you know, there are other dimensions as well.
41
00:06:44.729 --> 00:06:46.640
You know, micro-targeting of your, of your customers.
42
00:06:46.780 --> 00:06:59.640
You know, when, when you understand your customer need, when you understand your customer interactions, you can provide the right product, the right service at the right time without going for a full blast, uh, to, to, to everyone.
43
00:06:59.740 --> 00:07:09.600
Um, there are also optimization or productivity enhancement by bringing the right data to the right person at the right time.
44
00:07:10.380 --> 00:07:16.880
You know, if, if you think about, you know, a, a lot of knowledge graph, a lot of data fabric work, uh, that we're doing, th-that...
45
00:07:16.940 --> 00:07:36.160
those types of projects are typically about making sure that within a business process, you've got all the data you need to make the decision at the point you need to make it, instead of having to reach out to different team members, you know, um, maybe create an internal support request for someone to go and look back at a contract or find the right data in a si- in a data silo somewhere.
46
00:07:36.220 --> 00:07:45.420
So there's a lot of optimization there, which is not necessarily cost-cutting or anything. It's, it's about enabling people to do their job better. Yeah.
47
00:07:45.460 --> 00:08:00.500
It's, it's cutting down those silos, and the technology helps us better, well, combining those organizations where before you really had to put out that request and then unify it in, in, in your own way.
48
00:08:01.300 --> 00:08:10.310
Uh, MarkLogic, uh, the company you came from, which has now been acquired by Progress, is a leader in unifying that, uh, that data and integrating that data.
49
00:08:10.760 --> 00:08:24.120
Can you explain the unique approach that, that MarkLogic, uh, took, uh, on integrating that data? And what are the advantages now of being embedded in the, in the Progress, uh, data platform ecosystem? Sure.
50
00:08:25.220 --> 00:08:34.780
I mean, a-as you just mentioned, yeah, MarkLogic and the Progress data platform now, you know, it's like we're building around it. You know, we...
51
00:08:35.229 --> 00:08:47.130
Our objective here is to really unlock the value from complex data, you know, to make it, you know, easily discoverable, fit for purpose, you know, to build high-value, high-power applications,
52
00:08:48.180 --> 00:08:54.300
uh, nowadays to deliver decision intelligence and support your AI journey, of course. Uh, yeah, what do we do behind the scene?
53
00:08:54.340 --> 00:09:00.479
I mean, we integrate data wherever it, it is coming from, so, you know, breaking down those data silos.
54
00:09:00.520 --> 00:09:13.480
We're actually storing the part of the data we care about, so usually metadata, knowledge graph, entity, those things, et cetera. You know, we provide querying and, you know, retrieval capability on top of that.
55
00:09:14.580 --> 00:09:28.160
We have really powerful way to curate, govern, and secure data. And because, you know, storing data, managing data, and reaching data is one thing, uh, but using it is another.
56
00:09:28.600 --> 00:09:41.860
I mean, we have-- we are building and have built over the years, you know, really good capabilities around exploring and analyzing that data. So what are the, the pivot, the, the pillars of our technology?
57
00:09:41.900 --> 00:09:52.320
I mean, first of all, a truly multi-model database. You know, we can manage documents, graph, relational, geospatial, semantic, bitemporal data in a single engine.
58
00:09:52.360 --> 00:10:01.500
You know, we have really high performance search and query capability, so we can run complex queries across all of your data and get really accurate and sub-second results.
59
00:10:02.480 --> 00:10:14.140
We have powerful data integration and data management capability, so integrating data. We have that's, you know, the DataDirect part of the portfolio, which is part of the platform.
60
00:10:14.480 --> 00:10:25.320
Um, really impressive way of curating data and adding semantics and, you know, metadata management, uh, with Semaphore, um, agile modeling, et cetera.
61
00:10:25.460 --> 00:10:32.620
So we have smart mastering in MarkLogic as well as part of the Data Hub, um, and, and Data Hub Central product.
62
00:10:33.200 --> 00:10:45.140
So really powerful metadata and data management capability overall complemented by a comprehensive suite of tools for developers and business users. You know, why?
63
00:10:45.180 --> 00:10:54.160
Because, A, we want developers to build applications faster, um, to reduce the time to market. And why business users?
64
00:10:54.180 --> 00:11:04.440
Because as we mentioned, uh, as I mentioned already several times, you know, we want to put our tools, our capabilities in the hands of the people that run the business with their data.
65
00:11:05.420 --> 00:11:12.140
I mean, it, it, it's really important that a data project, it doesn't go at the pace of an IT project.
66
00:11:12.780 --> 00:11:22.120
We can power IT for everything to do with the integration, running the platform, but we need the data to be in the hands of the business users to move at the pace of the business.
67
00:11:24.500 --> 00:11:35.060
Are you referring that there is a different speed between IT and the business part? I mean, never heard about that, [laughs] but... No, no, I'm just, just laughing, but I do understand what, what you mean.
68
00:11:35.160 --> 00:11:47.600
So you provide as well with, with the Progress platform a way of businesses to easily, uh, create their, uh, data products and make data more accessible. Then you put it through IT.
69
00:11:47.740 --> 00:11:58.600
I mean, IT has its value by governing it, uh, by providing the architecture, making sure it runs twenty-four/seven, it's stable. I mean, that needs to be done.
70
00:11:58.790 --> 00:12:05.640
Uh, sometimes if you see it comes from the business, it's not so solid as it should be. It's not so secure as it should be.
71
00:12:05.740 --> 00:12:13.026
So that's the part where I see IT still needs to be-Knocking on the door and say, "Hey, guys, we need to slow down."
72
00:12:13.216 --> 00:12:22.696
Slowing down, not because of we do things so slow, but we need to do these, these things to make solid solutions in such a way. A-absolutely. Absolutely.
73
00:12:22.736 --> 00:12:32.196
I mean, the, the infrastructure part, you know, all the plumbing, the connections, the integration, all of those things, when those have to be governed, you know, go through a, a normal IT life cycle, et cetera, et cetera.
74
00:12:33.596 --> 00:12:46.776
Where many data projects break down today is when the data and the data interpretation is actually treated as a data- as a, you know, project life cycle artifact. I mean, data moves at a much faster pace.
75
00:12:46.826 --> 00:12:55.326
Its interpretation is different. We need... Essentially, we're bringing data agility. Kinda in, in the same way that, you know, engineering teams have been moving to an agile model,
76
00:12:56.916 --> 00:13:10.756
so you can deliver incremental functionality, you know, shippable incremental functionality at, at the end of this, of each sprint, and you're supported by a shared IT organization that makes sure infrastructure on which you run is, you know, secure, governed, et cetera.
77
00:13:11.596 --> 00:13:14.956
We want to do the same thing with data, all right? Build data agility.
78
00:13:15.596 --> 00:13:30.276
In-instead of engineering scrum team, you've got business scrum team, thinking about a data product, thinking about how they need to interpret the data, use the data in their operation, supported by an IT organization which provide them with the software, you know, and the infrastructure to operate in.
79
00:13:31.396 --> 00:13:57.776
Does that mean that the data platform from Progress and all the toolings around that now provide the two different modi where you can have the exploration part, which goes pretty fast on the speed of data, and that you can, once you have that data product developed or that vision, uh, for the data or the decision you wanna take with your data, and then move it through the, the slower path from IT, be fully governed, uh, fully secure, and you can easily switch between the two, uh, of those?
80
00:13:58.616 --> 00:14:16.736
I, I'd go even further than that because we completely separate the infrastructure part of the software, which is, you know, governed, et cetera, and goes through the IT life cycle as normal, from the data part, which is using a more information science, data science librarian-type, uh, workflow.
81
00:14:17.316 --> 00:14:29.076
So without changing the software configuration, you can change the modeling and an ontology and try a new data scenario, try a new data interpretation in a matter of hours, in a matter of minutes, in fact.
82
00:14:29.776 --> 00:14:47.396
Um, and, and not only that, but, you know, from a technology perspective, because it is all semantics and W3C standard compliant, you can actually have several of those interpretation running side by side, and even at query time, you can say, "Oh, I want to run through the lens of this world or that business domain or that process."
83
00:14:47.916 --> 00:14:59.846
So there's no, you know, hard-coded schemas. There's no hard-coded, you know, view of the world. There's no, you know, predefined, you know, data reduction. And I think data- Yeah. It's-...
84
00:14:59.856 --> 00:15:04.976
data reduction, loss of precision is actually something that, um, you know, we talk a lot about, um, in our project.
85
00:15:05.045 --> 00:15:15.446
'Cause earlier in the cycle, you do data interpretation, typically, the more, [sighs] the more information you lose. Yeah. It's, it's data agility.
86
00:15:15.856 --> 00:15:26.056
And when you say we, we can through-- look through the different lenses, we, we-- Well, I'm the traditional schooled data warehouse builder, dimensional modeling, uh, data guy.
87
00:15:26.636 --> 00:15:36.356
Uh, and we've b- had the discussion of one version of the truth where Bill Inmon, one of my former guests as well, uh, was very proud of, of providing that.
88
00:15:36.646 --> 00:15:50.846
But we've learned over time that there are different versions of the truth, depending on which department or where you are in, in your environment, and, and that's important that you can easily switch those views, but still on, on the same basis, the same foundation.
89
00:15:51.756 --> 00:16:02.896
Exactly. Because based on who you are, where you are, you might have varied different information needs, even though your query is the same. It might be based- Yeah...
90
00:16:02.906 --> 00:16:12.526
on access, it might be based on local translations, it might be based on, you know, regulations, all sorts of things. I mean, we, we do a lot of projects in the pharmaceutical space.
91
00:16:13.076 --> 00:16:24.236
Well, even if the research data is the same, even if the molecule, uh, is supposedly the same, it might not be manufactured in the same way because the recipe might change depending on the local ingredients.
92
00:16:24.246 --> 00:16:32.866
The brand might also be different depending on the market. The reporting requirements may be different. The little leaflets you get in the box may be different because that's based on local regulations.
93
00:16:33.326 --> 00:16:36.276
It doesn't mean the product is different. I mean, conceptually, it's treating the same thing.
94
00:16:37.156 --> 00:16:48.876
Um, people might not have the same symptom because symptoms are also something which varies depending on, you know, where in the world you live. So it's interesting. I mean, we, we could go on for, for hours on that.
95
00:16:48.976 --> 00:16:59.936
But if you, if you think, if you go back all the way to the root of ontologies and, you know, semiotic, not semantic, but semiotic, the relationship between the words we use, the objects in the real world,
96
00:17:00.896 --> 00:17:13.996
and the, the concept behind them, you know, th-that mapping varies drastically depending on audience. Yeah. F-fully understand what, what, what, what you're, uh, talking about.
97
00:17:14.316 --> 00:17:20.136
Now, with, with generative AI and the large language models coming in, and we're talking about different perspectives.
98
00:17:20.196 --> 00:17:35.616
I mean, people have used ChatGPT already for the last year, and where you s-- Well, there it's pretty clear that you can add a different role, and depending on the role you give it, uh, to the large language model, you get a different outcome.
99
00:17:35.866 --> 00:17:44.556
But telling you the same thing, but depending on, on the role you give it, you get a different wording. That's, that's exactly what, what you're referring to, I think.
100
00:17:45.376 --> 00:17:59.136
And how do you see these large language models will, will help, uh, revolutionizing, um, the, the space of data management? Will that help us in, in building better solutions, uh, more valuable solutions?
101
00:18:00.116 --> 00:18:04.436
I mean, it is profoundly changing the field of data management. I mean, no doubt about it.
102
00:18:04.776 --> 00:18:13.456
I mean, from a technical perspective, because it gives you new tools and new approaches, um, you know, new ways of interpreting a, a corpus, new way of serving a corpus.
103
00:18:14.596 --> 00:18:26.932
Um-From my perspective, it's also really interesting because it is driving a very different type of adoption for metadata and data management. I mean,
104
00:18:28.232 --> 00:18:38.342
a lot of customers, a lot of prospects are coming to us saying, "Hey, we've started this generative AI project, but it's not really performing the way we want." Why?
105
00:18:38.632 --> 00:18:46.622
Well, what have you given it as an input, you know, for, for RAG strategy, et cetera? It's like, well, the data we have. Well, guess where the problem lies, right?
106
00:18:46.792 --> 00:18:55.332
So, you know, from an adoption perspective, generative AI is really interesting because, you know, it is fueling the need for better data.
107
00:18:57.292 --> 00:19:08.332
Our customers, our prospects, our partners are all telling us, you know, of horror story where customers have tried to use their data as is to drive a generative AI solution. You know, same old, same old.
108
00:19:08.792 --> 00:19:16.512
You know, I'm not going to say garbage in, garbage out because actually, the generative AI solutions are making a good job at hiding some of that.
109
00:19:16.532 --> 00:19:25.812
But the reality is, if you want precise and accurate and trustworthy answers on private data, today, you need a RAG strategy.
110
00:19:25.852 --> 00:19:34.892
Today, you need, you know, the right level of security, the right level of understanding, the right level of search. You know, and that goes all the way back to our story of breaking the data silos.
111
00:19:35.712 --> 00:19:46.052
You know, moving from a application-centric architecture to data-centric architecture, having, you know, knowledge graphs, having a data fabric from which you can actually surface the right data at the right time.
112
00:19:46.652 --> 00:19:52.742
You know, and the conversation we just had about micro-targeting our users, about understanding our users' behavior,
113
00:19:53.792 --> 00:20:09.032
I mean, all of that goes right back in the way we search our fabric to include the right fragment in the RAG, you know, to support the generative AI response generation. So from that adoption perspective, we're seeing a
114
00:20:09.952 --> 00:20:15.372
renewed interest in data management and in metadata management and, and search technologies in general.
115
00:20:16.752 --> 00:20:24.352
The-- Associated with that, I suppose, you know, there's a more social or societal perspective, which is data is becoming cool again.
116
00:20:25.032 --> 00:20:33.532
You know, you, you hear of people, you know, trying things with ChatGPT, trying their own, you know, AI, ML pipelines and MLOps, and it, it's great actually.
117
00:20:33.572 --> 00:20:43.112
I mean, you know, socially, I'm now answering questions about, you know, stuff we do in our research labs, and it's actually really interesting. People didn't pay attention ten years ago.
118
00:20:43.552 --> 00:20:59.302
[laughs] However, uh, you know, you were asking about impact and changes. One big impact and change we're seeing actually in the way we interact, um, on a lot of data projects is a, a
119
00:21:00.492 --> 00:21:14.572
realization of the legal and threats associated with how data is used. So, you know, DLP, uh, privacy, all of those things are fueling actually a lot of additional inquiries.
120
00:21:15.212 --> 00:21:25.772
Um, you know, a lot of additional time spent on projects designing, uh, what should have been, you know, catered for all along. Uh, but certainly, um, organizations are now understanding
121
00:21:27.632 --> 00:21:34.642
that their data, the minute it's, you know, out in the LM may or may not belong to them, and they may or may not be in control, and there are unknowns.
122
00:21:34.832 --> 00:21:45.052
So we're seeing a lot of activity in, um, you know, sec- with security teams, with legal teams, with procurement teams around, you know, how do you handle, um, this new landscape?
123
00:21:47.172 --> 00:22:00.092
So what you're seeing here is that previously we had to pay attention to the governance and the security of the data, but because it was more siloed, it went lesser out, let's call it like that.
124
00:22:00.792 --> 00:22:09.692
Uh, people were not paying so much attention to it. I recall the days where you said, "Okay, I need access to the server, to the database. Here are the admin rights, full open, whatever you do with it."
125
00:22:10.612 --> 00:22:21.871
Uh, you could do as a developer kind of everything on, on the server, but you're very well aware of the security breaches of the privacy regulations and all of that.
126
00:22:21.932 --> 00:22:34.311
What you need to take into account, uh, and be careful with because you see so much, uh, valuable information, uh, which shouldn't be for all eyes, but there is a kind of trust towards developers to develop that product.
127
00:22:34.792 --> 00:22:50.272
But now with the large language models, this information is easier to go out of the organization, so people become more aware of, of governing that, of protecting, uh, the privacy of, of all the information which goes into the model.
128
00:22:50.492 --> 00:22:58.472
Uh, and that's where they now become more aware and paying more attention to, to that part of, of data management. Absolutely.
129
00:22:58.512 --> 00:23:09.672
So, you know, until relatively recently, the majority of security work associated with project was, you know, doing role-based control, compartment security, et cetera, within the database.
130
00:23:10.152 --> 00:23:18.882
Making sure that the data in the system or in the s- in, in the fabric, in the search interfaces, you know, we're reaching the right person.
131
00:23:19.592 --> 00:23:31.312
You know, now we're seeing a lot of organization go beyond that and either contractually or through systems, making sure that once that data is served in the application, there's actually traceability beyond that.
132
00:23:31.372 --> 00:23:36.972
If it's sent out to a customer, uh, if it's actually sh-shared with another coworker, you know, all of those sort of things.
133
00:23:37.052 --> 00:23:49.852
Um, and, and especially, you know, restrictions around using that data in training LLMs, in AI, um, or machine learning projects, et cetera. So, so you see,
134
00:23:50.832 --> 00:24:04.312
there's definitely a recognition now that using data for the purpose of training has risks. It doesn't mean we shouldn't do it. It means that we should understand the full scope
135
00:24:05.552 --> 00:24:19.652
and accept the risk when we want to accept the risk. So it's a different approach. It's-- I, I feel like when you say we, we want to... You have two s- two modi again in, in developing solutions.
136
00:24:19.732 --> 00:24:29.380
If you want to be in-innovative, creativeYou have to-- Well, you're building less secure and governance solutions because you wanna go fast, and that is slowing you down.
137
00:24:29.880 --> 00:24:41.260
And on the other hand, if you want to have it protected, that takes ages before you have the regulations in place and, and so on. So that's, that's a bit what I see where you have to find a trade-off.
138
00:24:41.600 --> 00:25:03.060
Do we use that data to take the risk to go faster to build really an innovative product as such, or do we really put the foot on the brake and, uh, within the next one hundred years, maybe we have the next [laughs] large language model, and not at the speed where you see every week you have a new version of a, of an open source large language model.
139
00:25:03.480 --> 00:25:15.900
Technology goes at such a fast pace. I mean, I've been diving into large language models now in-depth into for the last six months, and if you see, you take frameworks like, like LangChain, it changes every week.
140
00:25:15.960 --> 00:25:21.940
Documentation changes, fra- uh, the complete framework changes after a month, and then you kinda...
141
00:25:22.020 --> 00:25:28.500
Well, it's, it's hard to stay on top of it on the development and on everything what is on the market to build those solutions.
142
00:25:28.510 --> 00:25:42.140
And I think that's, that's gonna be a challenge now for, for most of the people to understand, one, on what is possible with the technology, but as well on capturing that in, that, that capability and using that to build the innovative products.
143
00:25:43.700 --> 00:25:57.840
So talking about that, how do you see the, the data management space evolving over the, the next to five, uh, five to ten years, and, and what emerging trends, uh, should data professionals and executives, uh, should be paying attention to?
144
00:26:00.900 --> 00:26:15.300
Well, data quality, of course, is, uh, a big one for me. Data management, right? [laughs] Metadata management as well. Um, so five to ten years, I mean, [sighs] you've seen the, the pace of change right now.
145
00:26:15.420 --> 00:26:20.160
I mean, uh, uh, in some of those topics, it's hard to predict what's going to come in two weeks. So, um,
146
00:26:21.080 --> 00:26:35.900
this said, and, and back to what we said a, you know, a few minutes ago, there's this very big dichotomy, you know, difference in, in swim lane, in pace between our industry and manufacturing and, you know, et cetera users, regardless of the, of the industries they're in,
147
00:26:36.880 --> 00:26:45.760
and our research. Uh, so if you look at projects in production or, you know, projects that are dealing with data at scale, I think we're still,
148
00:26:46.980 --> 00:26:50.800
you know, in the early days of data fabric and knowledge graph and, a-and all of those things.
149
00:26:50.820 --> 00:27:01.780
So there's still, you know, a lot of technology to be developed, a lot of technology, a lot of processes to be simplified, uh, to make it applicable and make sure it's rolled out and successful into the environment.
150
00:27:01.820 --> 00:27:13.400
I mean, we're still seeing the same three big hurdles, um, you know, for, for our customers, which is the plumbing, uh, the data quality, and actually operationalizing your data.
151
00:27:13.469 --> 00:27:21.860
So, you know, that's the, the order in which as an organization you need to, to, to jump, um, and, and the order in which you're, you're falling as well.
152
00:27:21.920 --> 00:27:38.200
And, and just something like plumbing, I mean, you just said it a minute ago. LLMs, new models coming out on a weekly basis, right? The modern data stack is fantastic. You know, it's great, it's composable.
153
00:27:38.240 --> 00:27:47.940
But for many, it is too composable, and they don't know where to go anymore. Um, you know, and, and that results in projects which are... You know, it's fantastic for research projects.
154
00:27:48.240 --> 00:27:56.940
For, for production projects, you know, what framework do I use? What technology? What API? I mean, it's changing on a weekly basis, on a daily basis in some cases. Hold on, I've, I've just come in this morning.
155
00:27:56.980 --> 00:28:08.240
My API, you know, my call doesn't work anymore. What, what's going on? So, you know, one thing that we're really working on, um, in, in the Progress data platform is making sure we've got an end-to-end story.
156
00:28:08.260 --> 00:28:10.680
Yeah, so you don't have to reinvent that plumbing.
157
00:28:10.740 --> 00:28:28.120
And, you know, if you think about the strengths of Semaphore, the strengths of MarkLogic, the strengths of DataDirect, the strengths of Corticon, you know, the strengths of Kendo, another Progress product that we're using for visualization, you know, we have all of the elements to really make high power, you know, high value applications.
158
00:28:29.500 --> 00:28:38.060
And the data platform strategy is essentially creating the blueprint. So as a customer, you don't have to reinvent the wheel. You can follow a proven recipe.
159
00:28:38.980 --> 00:28:45.440
So your data projects can be integrating with the data sources all the way to delivering value in a business application.
160
00:28:46.980 --> 00:28:55.620
Yeah, as f- as soon as possible, as fast as possible, minimum time to market, minimum complexity, you know, minimum investment in the commodity, in the plumbing.
161
00:28:56.300 --> 00:29:05.890
Because where the value lies from a business perspective is in the data quality. You understand your data. You are producing that data for a reason. So what is that reason? Can you model it?
162
00:29:06.640 --> 00:29:19.040
Can you actually identify it in content? And how can you leverage it to change the way you do business? You know, and, and very often, projects are, you know, [sighs]
163
00:29:20.940 --> 00:29:32.240
completely submerged in the quagmire of choosing the right moving API, you know, uh, or, and the right component for the day. Then they completely lose track of, of the rest of the data journey.
164
00:29:33.960 --> 00:29:44.880
Now, of course, I imagine there'll be some stabilization, some consolidation, and, you know, we're already starting to see that consolidation. Um, but we need to provide, I think, uh, as a,
165
00:29:46.500 --> 00:29:59.000
you know, as an industry, actually, some guidance on how to compose the right architecture to achieve the right data quality, so you can operationalize your data. All the while, maintaining security
166
00:30:00.800 --> 00:30:06.600
and helping you enforce governance. At the end of the day, without governance, you won't have trustworthy data.
167
00:30:07.560 --> 00:30:19.700
Without security, you have all sorts of breaches, repu-reputational issue, and you'll be caught up, you know, caught up in the crossfire of solving, you know, data security breaches rather than solving your business problems, right?
168
00:30:19.800 --> 00:30:34.442
So back to square one. [laughs]We, we have a lot of more possibilities because we have so much choice of technology, but it made it only more complex. Uh, so that's, that's my takeaway from what you just said.
169
00:30:35.112 --> 00:30:45.512
But as a, as a vendor, you're helping customers as well with the blueprints where you say, "This is what we've tried. We've put a lot of research in it. We know this solution works."
170
00:30:45.692 --> 00:30:53.132
Maybe you have so many other different vector databases, what is possible. We, we know for this type of problem, this is the best to use.
171
00:30:53.382 --> 00:30:58.152
This is the plumbing you can use, this is the visualization, and this is how you can build your product.
172
00:30:58.672 --> 00:31:11.012
If we see you have a different need, we can revisit and give you a different blueprint and then, uh, migrate from the, from the previous one to this newer one with your new, uh, requirements, uh, if I understand.
173
00:31:11.612 --> 00:31:22.592
Ab-absolutely. And one thing I'll, I'll probably jump on. You just said, "This is the best one. This is the best one. This is the best one." Well, actually, you don't necessarily need the best one.
174
00:31:23.472 --> 00:31:31.412
You know, sometimes the best one is actually too complicated, too expensive, too whatever, too hard to manage. I-it's the right one for the use case.
175
00:31:32.372 --> 00:31:59.132
You know, f-for example, we may not have the best vector database ever in MarkLogic, but because we vector as well as, yeah, all of the other data types that we support, all the other query types that we support, all of the other indexing strategy that we support, it's actually the best for the use case because it allows you not only to operationalize your data, do your, you know, analytics, but also serve your data through, um, you know, RAG for GenAI solution.
176
00:31:59.192 --> 00:32:12.492
So it's about how do we optimize the architecture and the product capabilities and make the right choices on the way to deliver an enterprise solution. Yeah. I mean, it's kind of flashback.
177
00:32:12.552 --> 00:32:24.372
I mean, I've just built a large language model, uh, solution as well, and I had so many choices, like you said, the vector databases, the transformers, the embeddings, uh, the all of that part.
178
00:32:24.452 --> 00:32:35.972
And because it was very new, I thought, okay, with what I read, I go for just a SQLite database, which is easy accessible. You can look at your data. You can see the metadata, how it works. Come on.
179
00:32:36.032 --> 00:32:40.752
If you didn't use MarkLogic? [laughs] It's too expensive. No. [laughs] Shocking.
180
00:32:40.872 --> 00:32:48.852
Uh, I mean, I think, I think it's a, it's a good use case, definitely, uh, to, to look at MarkLogic with the semantics and, and, and put that in place.
181
00:32:48.972 --> 00:33:03.412
But you're absolutely right where I said, okay, this is what I feel is the best one, and I know, depending on once I have the full solution built, I can revisit all the individual components, and I know where we can optimize if the outcome is not what I'm expecting.
182
00:33:03.472 --> 00:33:11.072
But first I need to build the solution with the individual components and then, uh, value, uh, evaluate that, that part.
183
00:33:11.792 --> 00:33:24.372
So looking at that on, on a more personal note, what values guide you in your personal journey and, and, and how does these influence your vision for the future, uh, for data and analytics?
184
00:33:26.372 --> 00:33:38.952
Well, I think I, I opened with that already. You know, it's making data a-and the knowledge within that data accessible so you can act in informed ways. I mean, we could call that democratizing data management.
185
00:33:40.481 --> 00:33:44.182
Um, you know, it's leveraging your data. That's, you know,
186
00:33:45.392 --> 00:33:55.292
what I've actually tried to achieve and what I've actually worked on in pretty much every single one of my roles since, you know, for the last twenty-five years, essentially.
187
00:33:55.612 --> 00:34:07.472
[laughs] And how does that influence, you know, my, my vision of the future? Well, clearly one-sided and fairly focused on the, on the data management and analytics, but, um, you know, it's about use case.
188
00:34:08.252 --> 00:34:23.592
It-it's about identifying high-value use cases, identifying where data can help. Um, yeah, because if, if you look at the way we operate today, eighty percent of data is unstructured.
189
00:34:25.512 --> 00:34:34.562
The vast majority of data that organizations are producing, exchanging is not used for decision-making.
190
00:34:35.132 --> 00:34:47.232
The, the twenty percent that lives in a relational world, that's what we put on BI dashboards, that's what, you know, our management teams are looking at on their weekly stats and, and scorecards. It's so...
191
00:34:47.292 --> 00:34:57.591
such a small fragment of the overall data landscape we've got to actually make decisions. So, you know, changing that status quo i-is really the motivator here.
192
00:34:57.632 --> 00:35:11.512
You know, and can we bring the full context at the point of decision? And the full context comes from understanding, you know, the full scope of the data and interpreting it, it in the right way.
193
00:35:11.552 --> 00:35:20.442
So, you know, going back to semantics, that allows you to understand the context of the data from, you know, the point of origin. You know, why was it created in, in that way?
194
00:35:21.312 --> 00:35:28.102
And, and some of it is technical, absolutely, and that's why we've got, you know, all the, all the product suites that we have to help you in that data journey.
195
00:35:28.572 --> 00:35:34.452
But a lot of it is also, you know, data science, and a lot of, of it is also library science.
196
00:35:34.512 --> 00:35:44.652
I mean, good old traditional looking at data, you know, talking to the data producers, understanding what they mean, what they, what they've tried to do, why they did it.
197
00:35:45.592 --> 00:35:58.752
And, and sometimes, you know, even today, you-you've got some-sometimes absolutely fantastic answers. It's a, it's a critical mindset. It's, it's that challenging and, and just what you see, not always believing.
198
00:35:59.012 --> 00:36:09.932
I, I see that as well with, we, we have that with the ChatGPT, is you, you put, you, you put a question forward, you get an answer, but you always need to challenge that answer. Is it really right?
199
00:36:10.452 --> 00:36:21.582
And you need to be sure that it is factual correct in, in such a way. So that, that demands a, a critical mindset of, of how you look at your data and, and how you take that information.
200
00:36:21.702 --> 00:36:33.652
It's like the journalists back in the days, they, they were forced to just double, triple verify all the information they got. But now with the speed or at which they need to workIt's, it's getting upended.
201
00:36:33.752 --> 00:36:40.202
Sometimes they get a fact and they think it's true, and give that out, and then don't do the, the double fact-checking.
202
00:36:40.672 --> 00:36:54.112
Now thinking about that, what advice would you give to newcomers that are entering into the data and analytics space and, and make that really their business, their, their roadmap to, to be successful? Be curious.
203
00:36:56.172 --> 00:37:04.102
Yeah, curiosity is actually the number one skill here. You know, because, you know, we often talk about data mining,
204
00:37:05.152 --> 00:37:14.242
you know, as, you know, one of the starting points of creating your data journey, 'cause you need to understand what you have and what it means and, and all those things. You know, you need to be curious about
205
00:37:15.512 --> 00:37:23.092
how data came to be. You know, why was it produced? Why was it produced the way it is? You know, what was meant by the author? What is needed by the audience?
206
00:37:23.472 --> 00:37:28.001
You know, some of it is, you know, business acumen, some of it is technical, some of it is...
207
00:37:28.262 --> 00:37:38.452
You know, the bottom line is if I look at all of my teams, the successful teams are those that are not losing the curiosity over time. You know, keep trying to discover. You know, it's a--
208
00:37:39.632 --> 00:37:52.332
Ultimately, you know, data management is detective work, and you need to be on your feet, you know, looking, scrutinizing all the time. Yeah, everyone could be a data citizen. Everyone is a data citizen.
209
00:37:52.912 --> 00:38:04.332
You know, some people might not know it yet. It's only a matter of time. [laughs] But do need-- do organizations need to provide a bit more time for people to allow that exploration?
210
00:38:04.592 --> 00:38:14.272
Uh, like you say, you say, "Be curious," which is very general, but then you add it onto it, it can be technical, it can be business, it can be data-related.
211
00:38:14.312 --> 00:38:30.252
So it's not only in the domain where you are or are a specialist, but be more holistic in, uh, in being curious across the business where you're working in, uh, in such a way. Absolutely. And, you know, some of the more,
212
00:38:31.952 --> 00:38:42.192
I was gonna say interesting, some of the more successful, uh, projects that we've seen, uh, were completely, you know, organized in ways that are non-traditional.
213
00:38:42.752 --> 00:39:02.312
You know, we, we had a, a really interesting project which started as a almost one-man show, uh, from a, a doctor in a line of business, uh, that was actually a bit puzzled by some of the data that he was observing, and he started building models, you know, doing things in R and doing, you know, doing all sorts of things.
214
00:39:02.352 --> 00:39:10.032
And we, we met him, you know, in, in one of the interview at that client. The client didn't even know that they had such a, a doctor in the organization that was looking at data.
215
00:39:10.052 --> 00:39:15.372
We had some fantastic conversation, and he became one of the data stewards for, you know, one of the, one of the big project.
216
00:39:15.592 --> 00:39:35.532
Brought his experience of the field, brought his experience of, you know, what things mean in the context and how you com- you combine what the data says with the clinical science, with your interpretation of how you interact with a patient to actually do essentially data science, but with the human in the loop and an understanding of the human side of it.
217
00:39:36.212 --> 00:39:44.632
You know, a-and I think that's what I mean by curiosity. Always keep all of the variables, all, all of the dimensions of the project, you know, in mind
218
00:39:45.652 --> 00:39:49.982
when you're, you know, when you're designing solutions, when you're thinking about the results, when you're thinking about the algorithms.
219
00:39:50.032 --> 00:39:59.591
I mean, if you go at it from a purely technical angle, you might get something interesting, but is it true? Is it relevant? Is it going to make a difference? Yeah.
220
00:40:00.182 --> 00:40:13.512
Stay interested, stay motivated, and stay energized and, and, and you will find your, uh, answers of, of what you're looking for. Uh, Mathieu, we-we're coming to the end of, of the show. Uh, I always close up with
221
00:40:15.052 --> 00:40:28.091
data is connecting us all, but music is connecting us all as well, so I would love to hear what is your favorite band or favorite type of music? It's a, it's a very hard question, 'cause I've got quite an eclectic taste.
222
00:40:28.152 --> 00:40:38.162
So I'll, I'll give you a f- a few elements of answer. So I'm actually a, a saxophone player. I don't play a lot. Uh, whenever I got time, I, I, I try to do a, a little bit still.
223
00:40:38.792 --> 00:40:54.732
Uh, right now I've got couple of scores on my, uh, you know, on my stand, uh, one from Paolo Conte and one from, uh, Astor Piazz- uh, Piazzolla, so Cafe 1930. Uh, really interesting. Love them.
224
00:40:55.012 --> 00:41:03.352
Uh, this said, you know, on the other end of the scale, uh, tonight is actually the Brave Festival, and we're going with the family to, to watch Sting. So [laughs] there you go.
225
00:41:03.372 --> 00:41:15.072
[laughs] Oh, he's still performing, so that's interesting to h- to hear. Yeah, yeah, yeah. That's, that's quite a very, uh, va- uh, variation on taste of music.
226
00:41:15.132 --> 00:41:23.932
But, uh, really, uh, ex-- well, discovered a lot of, uh, music with my guests which I've never heard before, so this is the fun part for me to close out.
227
00:41:24.312 --> 00:41:37.752
Uh, Mathieu, it was so, uh, much fun having you and interesting, uh, on the Data Gurus, uh, Strategy Podcast. So thank you for coming and having this, uh, very nice chat. Thank you for having me. Always a pleasure.
228
00:41:37.812 --> 00:41:48.032
Thank you for joining us on this awesome podcast. As senior executives, data and analytics architects, and AI professionals, your time is valuable and we appreciate you choosing to hang out with us.
229
00:41:48.072 --> 00:41:54.552
If you liked what you heard, please give us a thumbs up, hit the subscribe button, and leave a comment. We love hearing from our audience.
230
00:41:54.592 --> 00:42:10.091
Don't forget to spread the word on social media, and let's continue to drive innovation in the industry together. Thanks for listening, and we'll catch you on the next episode. [outro music]
