Bản chép lời đầy đủ của Unlocking AI's Potential Through Data Readiness

and repeat all the way back to, maybe, step one okay, we have to redefine our problem or our use case, or is it two? 00:02:28.324 --> 00:03:01.181 So you can iterate through any of those steps, tweaking all of the elements here to really figure out and getting your optimum outcome, and this is a challenge, just I see, facing a lot of companies, and as you go through this, there's going to be breakdowns, and one of the things that I kind of thought was maybe interesting to talk about today was the idea about data readiness, because I see that as being one of the most challenging issues that organizations really struggle with. 00:03:01.181 --> 00:03:03.084 How do you get the data structured? 00:03:03.084 --> 00:03:04.265 Where is it? 00:03:04.265 --> 00:03:05.407 Are we getting everything we need? 00:03:05.407 --> 00:03:08.413 Are we getting the right information and the right data? 00:03:08.413 --> 00:03:14.752 And really we see that this both applies to both generative AI as well as agentic AI projects. 00:03:15.560 --> 00:03:19.371 You know, and sometimes you know you and I have talked in the past about the data value chain. 00:03:19.371 --> 00:03:21.544 Do you want to talk a little bit about that? 00:03:21.544 --> 00:03:23.849 I think it kind of lines us up as well for this discussion. 00:03:24.570 --> 00:03:26.640 Sure, thank you for asking that. 00:03:26.640 --> 00:03:33.633 So the data value chain really goes through and it really becomes part of the whole AI project. 00:03:33.633 --> 00:03:38.731 Because one is how do you capture, how do you identify this data? 00:03:38.731 --> 00:03:46.834 And, as we're going through it, every organization has just been storing all of their files, all of their data for years. 00:03:46.834 --> 00:03:50.825 There's value to it, right, there's value to the organization. 00:03:50.825 --> 00:03:53.120 If there wasn't value, why are you storing it, right? 00:03:53.280 --> 00:04:06.936 So, with that, how do you start to identify, how can we convert this from just being able to store it and start to create it as an asset that actually can be monetized or has an ROI component? 00:04:07.740 --> 00:04:32.012 And so, as you're going through this and applying either machine learning or AI elements on this, you're creating inference, you're creating outcomes to better understand the data, so you can then have strategic value of how to create automation, how to create analysis, look at how to do for some different modeling and predictions. 00:04:33.040 --> 00:04:48.613 So, with that, the data itself, those files, have the ability then to keep with it metadata, information, more about it that enriches it and starts to really tell the story about what this file actually can mean. 00:04:48.613 --> 00:05:09.451 And then how it can be applied is where we start to look at OK, the data analytics, the data insights and how that applies, and then can create value for the organization, and so it's an interesting concept because everyone has all of the data, but it's how you go through that process step to really create that value chain. 00:05:09.451 --> 00:05:27.663 One of the challenges around doing that process is how do we know that we have everything that's relevant over years, 20, 30 years? 00:05:27.663 --> 00:05:35.322 I have just silos and silos and silos of different data that we are either actively using or have actively used in the past that has a lot of intrinsic value that can be pulled from it. 00:05:35.322 --> 00:05:38.872 How do you make sure that you can easily get access to those? 00:05:38.872 --> 00:05:43.432 And so I see that that becomes a challenge a lot of times to a lot of larger organizations. 00:05:44.240 --> 00:05:44.862 This is perfect. 00:05:44.862 --> 00:06:01.161 So, talking about the data silos, and you know, with the acquisitions that companies have, or just becoming global in the multiple sites that they have, it's really easy to see how quickly you know this data can grow and how complicated your environment can get, and so you know. 00:06:01.161 --> 00:06:10.696 It leads us right into what we want to talk about today with the problem statement is getting back to AI, and why do organizations struggle to get started with AI adoption? 00:06:10.696 --> 00:06:18.232 And it's really with this core issue we're talking about why is it so difficult for companies to get their data ready for AI? 00:06:18.232 --> 00:06:19.985 So tell me a little bit about that. 00:06:20.199 --> 00:06:24.560 Yeah, it's really interesting because AI has this big mysticism. 00:06:24.560 --> 00:06:26.841 Right, it's this shiny thing Also it's really interesting. 00:06:26.882 --> 00:06:28.742 I mean because AI has this big mysticism, right, it's this, ooh, it's shiny thing. 00:06:28.742 --> 00:06:29.182 Also, it's really cool. 00:06:29.182 --> 00:06:36.365 Everyone wants like we need AI, we want to do it, we know we need it, okay, so what, then what? 00:06:36.365 --> 00:06:45.767 But we're finding what's interesting is, once you kind of get your mind around it, the biggest challenge isn't AI itself or being able to implement the AI project itself. 00:06:45.767 --> 00:06:50.649 It's to make sure you've got the right data that you're feeding into it. 00:06:50.649 --> 00:06:54.009 So you know that whole idea about data readiness. 00:06:54.951 --> 00:06:59.391 So for an enterprise business, it's typically a high stakes environment. 00:06:59.391 --> 00:07:11.055 There could be a lot of great and good things that come up from it, but there can also be a lot of caveats and a lot of issues and challenges and possibly mistakes that you want to make sure that you're aware of. 00:07:11.055 --> 00:07:22.119 So as you go into it, you've got your eyes wide open, but we see that at the heart of any AI project cycle, of those five steps, it really comes down to three key things. 00:07:22.119 --> 00:07:26.824 First off is what is the problem you're trying to solve? 00:07:26.824 -->

00:07:30.190 If you don't know the problem, you're not going to get a good output. 00:07:30.271 --> 00:07:32.504 It's this whole idea about when you prompt an AI. 00:07:32.504 --> 00:07:40.451 The more clear your prompt is, the better the outcome of what you're getting, and that's really really clear and really important here. 00:07:40.451 --> 00:07:44.829 The next thing is what are the data sets that you're going to use? 00:07:44.829 --> 00:07:51.093 And, within that data set, how do you know that it's everything you need in order to be successful? 00:07:51.093 --> 00:07:54.182 And, more importantly, are we missing anything? 00:07:54.182 --> 00:08:04.651 Or are we just doing too much, because you can throw up everything and then realize, well, we only needed maybe 50%? 00:08:04.711 --> 00:08:05.473 of what we did. 00:08:05.473 --> 00:08:08.689 So it's really the idea of is this what we needed? 00:08:09.019 --> 00:08:10.165 Is this the right data set? 00:08:10.165 --> 00:08:12.408 How do we curate and start to get our arms around it? 00:08:12.408 --> 00:08:17.851 And it's easier when you're smaller and have less data to go through. 00:08:17.851 --> 00:08:21.939 But you start compiling it and it just compounds. 00:08:21.939 --> 00:08:31.721 Once you start getting multiple systems, multiple disparate locations and different types of technologies, it can get really challenging to figure out how to go through this. 00:08:31.721 --> 00:08:47.705 Then the other important piece is probably the last piece of this is then how are you going to monitor, how are you going to test the outcomes and then be able to make the changes, to iterate and tweak it so you can get to that final optimum outcome that you're looking for? 00:08:47.705 --> 00:08:52.466 So what's interesting I found is that I've been researching this now for a long time. 00:08:52.888 --> 00:08:58.849 Probably last long time meeting, you know, in AI project world is probably four months, maybe six months. 00:08:58.849 --> 00:09:27.322 But what I find is that you know in all recent studies going back and this is articles from Forbes, ibm, different consulting groups that are published on the web that 42% across the board have identified that on their polls, all the respondents identified that data availability and data quality has been identified as one of the top five challenges for implementing a successful AI project and usually it's within the top three, depending on who you talk to. 00:09:27.322 --> 00:09:36.106 Once you start to add then transparency, data governance, oversight now combining those together, it's 100%. 00:09:36.106 --> 00:09:50.302 These are the challenges that everyone's dealing with and that, if you don't have a good answer or a good approach for, you're not going to have a successful outcome for your AI project and the result of that is going to be lost. 00:09:50.302 --> 00:09:52.283 Time could be lost. 00:09:52.283 --> 00:10:10.609 Does a good behooves any company to create a successful plan for a successful implementation? 00:10:11.692 --> 00:10:12.214 Makes sense. 00:10:12.214 --> 00:10:14.120 So you know many organizations. 00:10:14.120 --> 00:10:27.554 They struggle because they have old repositories it may be hard to access that repository, or they have in the cloud and their struggle to be able to utilize that if they're trying to pull that on a daily basis or many other potential challenges, and so to be in the cloud and they're struggling to be able to utilize that, if they're trying to pull that on a daily basis or many other potential challenges. 00:10:27.554 --> 00:10:31.931 And so, to be clear, it's AI isn't the roadblock itself in any way. 00:10:31.931 --> 00:10:40.375 It's the data disorganization or silos, or access and understanding of what data they have. 00:10:40.375 --> 00:10:49.475 When you think about this holistically, that companies have to get their hands around before AI can really be effective and prove the value that it's promising. 00:10:49.475 --> 00:10:52.206 So cleaning up your house per se right? 00:10:52.948 --> 00:10:57.130 Basically, yeah, because if you look at it I mean, data is the foundation of any AI system. 00:10:57.130 --> 00:11:03.490 If you don't have it, you're not going to get anything out of it, and it doesn't matter if it's structured, unstructured third party. 00:11:03.490 --> 00:11:15.243 But one of the biggest things we see that is a challenge is dealing with that unstructured data. 00:11:15.243 --> 00:11:22.727 The big data Companies have been buying a bunch of storage, whether it be cloud libraries, archives, all silos, legacy systems and AI projects, the technology. 00:11:22.727 --> 00:11:36.236 I'm just amazed at how fast it's been developing and it really has gotten to the point where it has potential to deliver some incredible insights, incredible automation and some of the new things that are coming out with the agents. 00:11:36.236 --> 00:11:37.557 I'm blown away with. 00:11:37.557 --> 00:11:48.412 I mean, I remember it would take us months to do something that now, with like five, ten lines of code, you can come up with the same result. 00:11:48.412 --> 00:11:57.181 So it's going to have a very significant effect on how we're going to use technology to align this to how we operate and how we operate on a daily basis within a company and delivering value to customers. 00:11:57.181 --> 00:12:08.004 And it comes back to if that company doesn't have access to that clean, well-organized data set, then you're going to be sitting trying to figure out how to deal with all the fragmented records. 00:12:08.004 --> 00:12:09.688 How do we get access to this? 00:12:09.688 --> 00:12:19.152 And it can get costly just for that aspect of it, and that's one of the things that we really identify this, what we could provide value and be of help and service with. 00:12:19.614 --> 00:12:30.230 We also identified that another significant challenge that customers are facing

is how do you integrate that data and the data sources, the data from various different sources. 00:12:30.230 --> 00:12:39.143 So, again, you know, silo data stored in disparate systems, making that difficult to compile and create a cohesive data set. 00:12:39.143 --> 00:12:47.764 So then you start adding IoT devices, you start adding remote systems, remote sites, and all that starts to add just the complexity of this data integration. 00:12:47.764 --> 00:12:53.216 And so some organizations will say, well, just give us everything and then we can sort it out later. 00:12:53.216 --> 00:13:00.933 Well, you could do that, but it's just kind of like taking well, hey, you know, I want to clean up my house by buying another house and moving all our stuff over. 00:13:00.933 --> 00:13:04.309 You end up just with the exact the same problem. 00:13:05.440 --> 00:13:07.625 So the issue with data quality alone. 00:13:07.625 --> 00:13:17.932 You start to see a cost develop from this and this starts to then, just alone, you start to identify well, jesus, we're going through this. 00:13:17.932 --> 00:13:19.907 How much revenue have we been losing? 00:13:19.907 --> 00:13:22.349 How many operational setbacks have we been having? 00:13:22.349 --> 00:13:25.389 How have we had to do all these gyrations to make this work? 00:13:25.389 --> 00:13:40.650 So, and you know just that alone, not even including AI, it's estimated that enterprise companies are typically losing probably around 10 to 12 million annually just due to poor quality of data and how they're accessing it. 00:13:41.032 --> 00:13:50.745 So kind of cool, by taking this project, an AI project actually gives us an opportunity to go back and clean the house, to really start to make things more effective. 00:13:50.745 --> 00:14:04.315 And so we see that with Crystalonix, we have the ability to significantly improve the outcome, not just for the AI project cycle, but also just for the organization in general. 00:14:04.315 --> 00:14:19.897 And our value-add proposition that we're delivering is that we know we can ensure data quality, be able to significantly decrease the time and cost it takes to do this and enable the creation of the data value chain. 00:14:19.897 --> 00:14:41.009 So if you look at just that, you know decrease in time and cost if we go through and take everything and upload it to the cloud I don't care what provider you want to use You're basically going through and saying, okay, team, go find it, copy everything, move it up there and then we'll figure it out Kind of kicking the can down the road in some respects. 00:14:41.770 --> 00:15:04.508 For sure, and I know some of the numbers that we've talked about are quite staggering with regards to what people are paying today or willing to tolerate with regards to time, if they're utilizing, let's say, like a third-party company, or some additional resources to try to sort through their data problem, to try to pull this together, and it's just super prohibitive. 00:15:04.508 --> 00:15:13.484 You know the timeline and the cost just to understand what data we have before we can even get to this point is it's crazy staggering. 00:15:13.484 --> 00:15:52.673 Love talking with you about some of the different and looking on your website and reading about some of the companies that AI adoption just really streamline this approach, get them there very quickly, potentially from some of the things we've discussed, over 10 to 12 months of data assessment in organization down to maybe even weeks for certain organizations. 00:15:52.673 --> 00:15:53.902 So how Talk? 00:15:53.902 --> 00:15:56.269 How does Crystal Onyx eliminate these obstacles? 00:15:57.520 --> 00:16:01.591 Well it really comes down to because we cut our teeth on this early on. 00:16:01.591 --> 00:16:18.029 Working with high-performance computing centers, we saw that there was going to be a challenge that in these organizations it was kind of a precursor to AI and you had all of these different silos of data that never did not talk to each other. 00:16:18.029 --> 00:16:27.504 So you had to have different systems be able to pull out, find something, transfer that over to another system that would then transfer it to another system. 00:16:27.504 --> 00:16:34.227 So it was really really not complex, but it was very manual, right, and you start doing this across large amounts of data sets. 00:16:34.227 --> 00:16:35.692 That just becomes very unruly. 00:16:35.692 --> 00:16:41.207 So we kind of got an early view of what we saw was going to be coming. 00:16:41.207 --> 00:16:48.490 And we worked with organizations like Bosworth doing AI autonomous vehicle testing. 00:16:48.490 --> 00:17:04.390 We saw the German Climate Center, which is the second largest supercomputing center in Europe, and how they were being challenged with all of the different types of storage environments and what they were currently trying to do to make it work. 00:17:04.390 --> 00:17:09.853 So we were able to come in and, first off, just be able to scan everything. 00:17:09.853 --> 00:17:32.090 So what's great about this is we saw this then applies directly to Advantage within an AI project for a company is we can go to any storage system that they have and scan it very quickly meaning that we're doing it a million or a million plus files per hour and start to create a global namespace. 00:17:32.090 --> 00:17:38.347 That global namespace is just a virtual representation of everything that we're seeing within their environment. 00:17:38.347 --> 00:17:40.772 So nothing's moved, nothing's touched. 00:17:40.772 --> 00:17:43.188 We now we just have a virtualization of it. 00:17:43.188 --> 00:17:47.984 Once that's done, we know everything about the files, we know everything about the storage. 00:17:47.984 --> 00:17:50.306 It's all there and you can start running reports. 00:17:50.306 --> 00:18:00.622 You're able to then query and, using our intelligent query system, start to

then analyze and identify everything that you have without having to ever have to touch the files. 00:18:00.622 --> 00:18:07.705 So what's great about that is it's not just a one-time shot, so it's not like we're just going and uploading, adjusting to somewhere. 00:18:07.705 --> 00:18:16.703 The system actually goes back and, based upon the business rules set by the customer or by the company Is it once every hour, is it once every day? 00:18:16.703 --> 00:18:23.685 And it's going through and just updating to making sure that we have all the latest, greatest files and everything that's updated. 00:18:23.685 --> 00:18:26.619 So you're continuing to work with the live model. 00:18:26.619 --> 00:18:51.148 The great thing about that is we can start using custom metadata, start to then go and create curation sets of everything that's there so what's primary and really customize it and start to break down that data set into what's the primary, what's most critical, what's the second level critical, what's third level critical, and start even creating iteration groups. 00:18:51.148 --> 00:18:55.105 So when you're testing you can do A-B testing, what's our outcomes. 00:18:55.105 --> 00:18:58.641 But it's a really easy way to do this and quickly. 00:18:59.241 --> 00:19:03.386 And so in our customers that we've done this with, where they were trying to go through and do it themselves. 00:19:03.386 --> 00:19:08.855 One great example use case is a customer as you mentioned or alluded to earlier. 00:19:08.855 --> 00:19:13.718 They went through and they had around three petabytes they were trying to create the data set on. 00:19:13.718 --> 00:19:15.767 It took them about 10 to 12 months. 00:19:15.767 --> 00:19:18.086 They don't know really what they got. 00:19:18.086 --> 00:19:23.867 They got some On their second pass, different project, but using the same data set. 00:19:23.867 --> 00:19:32.192 We were able to go through, get a completely everything and in fact we found that they were actually missing files because they couldn't get access to them. 00:19:32.192 --> 00:19:34.847 So we found another petabyte worth. 00:19:34.847 --> 00:19:42.767 So we actually had four that were able to go through, do this, get a complete, curated data set and do it within eight weeks. 00:19:42.767 --> 00:19:53.080 And then their whole thing went from 10 to 12 months with I don't know how many there's like 12 people involved, down to eight weeks with three people. 00:19:53.682 --> 00:19:55.765 So just that cost alone. 00:19:55.765 --> 00:19:58.871 What does that mean on a successful impact? 00:19:58.871 --> 00:20:08.586 And then does that mean that, gosh, we could even do two or three or four different projects within the same time period that otherwise it would take us to do one right? 00:20:08.586 --> 00:20:10.131 So how much more impactful can we be? 00:20:10.131 --> 00:20:22.365 So you know, it's taking a different approach, it's thinking outside the box and it's realizing that if I have all the data, I don't need to necessarily work with all those files and create all this translation and do everything that we need to. 00:20:22.365 --> 00:20:36.164 First let's just actually take a snapshot of it, virtualize it and then work with the virtualization, the reporting, the queries, take that curation set, get to what we need to and then take our next step, which is okay. 00:20:36.164 --> 00:20:42.085 Let's approach the modeling, and then we also enable that too, but that's a separate approach. 00:20:42.085 --> 00:20:46.525 The first thing, that's the most important, is just getting your arms around this data readiness. 00:20:47.226 --> 00:20:49.391 It's just a huge shift, right? 00:20:49.391 --> 00:21:00.226 I just want to reiterate that's just a huge shift to go from an experience where people are usually doing this manually, where they've pulled in 10, 15, 20 people sequestered to kind of manage. 00:21:00.226 --> 00:21:10.948 That's what I typically see whenever there's kind of a project where they need to migrate to a new tool or a bunch of data, move a bunch of data, or they're acquiring a company or whatever this needs to be for anything. 00:21:10.948 --> 00:21:19.007 They just grab, they sequester a large amount of people and it takes months for them to go through this sort this stuff. 00:21:19.007 --> 00:21:32.294 And so what I hear you saying again is that you can take an example and you've lived this from a 10 to 12 month manual scenario down to eight weeks with just a couple of people managing this process. 00:21:32.779 --> 00:21:36.611 And it's unheard of right how fast you can do it if you have the right tool sets. 00:21:37.059 --> 00:21:41.167 These are the kind of efficiencies that organizations are really searching for, right? 00:21:41.167 --> 00:21:48.816 We spend so much more effort for much less of an improvement, so this one, to me, is just quite grand in scale. 00:21:48.816 --> 00:21:51.585 So I'm just really excited that we're talking about this one. 00:21:51.585 --> 00:21:52.406 Yeah, thank you. 00:21:52.527 --> 00:22:02.234 Thank you and really appreciate the opportunity, you know, and invitation, and what's cool about it is, you know, even though it's a new tool, you're not actually changing anything. 00:22:02.234 --> 00:22:09.520 We're not saying, oh, you've got to do something completely different, it's just going away a slightly different way about it, and so that's. 00:22:09.520 --> 00:22:18.605 What's really nice is that we're not forcing anybody to change the way they're going to do business or the way they're going to, you know, change their organization or their approach. 00:22:18.605 --> 00:22:20.346 We're just going to help make it easier. 00:22:20.346 --> 00:22:33.194 And what we've found is, because Crystal Onyx is able to connect all the dots you know, and we create the visualization abstraction, we start to enable that flow of data, and this extends across. 00:22:33.734 -->

00:22:34.695 You know all these silos. 00:22:34.695 --> 00:22:43.593 So you've got your traditional NFS and SMB data shares NFS for Linux and SMB, typically for Mac environments and Windows environments. 00:22:43.593 --> 00:22:51.314 You've got all the legacy storage environments that go back years, even back to mainframes, also parallel file systems. 00:22:51.314 --> 00:22:54.644 So you've got Lustre, gpfs. 00:22:54.644 --> 00:23:05.815 These are all very, very fast file systems that are usually used for supercomputing and for AI and also for any tier zero where you've got your critical storage. 00:23:06.396 --> 00:23:14.314 Then we also even extend across to S3, object stores, clouds, multi-clouds and even able to manage tape libraries directly. 00:23:14.314 --> 00:23:29.174 We had one customer where we were able to go in and, over the course of four days, migrate the access of their entire 350 petabyte library archive and make it available for an open system. 00:23:29.174 --> 00:23:34.172 So it was going from a legacy environment to an open system environment and it took us four days. 00:23:34.172 --> 00:23:38.391 Everyone else was coming in and saying that's going to take 18 months, 20 months. 00:23:38.391 --> 00:23:46.788 It's like, well, no, we can just think about it differently and, using what we can do with the virtualization, we can make it available to you within four days. 00:23:47.490 --> 00:23:54.079 So those are really kind of the cool things that we can get involved with and really help make it successful and impactful. 00:23:54.079 --> 00:23:58.887 Difference and the thing I like about it is this just is not a one-stop thing. 00:23:58.887 --> 00:24:00.711 You don't just upload it and do it. 00:24:00.711 --> 00:24:02.313 You keep everything where it is. 00:24:02.313 --> 00:24:03.740 You keep doing your business. 00:24:03.740 --> 00:24:07.990 We'll come in and just do a scan, and it's a touchless scan. 00:24:08.171 --> 00:24:15.921 It's very easy and it's deployed within the environment, so you don't have to rely on a third party or external. 00:24:15.921 --> 00:24:34.310 This can be deployed within a company, within their data center, in their environment, either on a bare metal server, a VM environment or even, if you need to, a cloud compute environment or all of the above and all work in tandem together. 00:24:34.310 --> 00:24:39.721 So it's so flexible and because it's a multi-threaded, multi-process system, that's really really the key in this. 00:24:39.721 --> 00:25:01.099 Because it's so flexible and because it's a multi-threaded, multi-process system, that's really really the key in this, because it's very fast and we're doing everything in parallel, so it's not just doing one thing at a time, it's doing 10, 15 different processes all at a time and, as I mentioned earlier, typically we see a three-node cluster will often go out and be able to do this at a million or more files an hour. 00:25:01.645 --> 00:25:08.838 What I

love, what you just said, and I think it's worth reiterating, is that Crystal Onyx gives people the ability to bring this in-house. 00:25:08.838 --> 00:25:12.756 It's the utility that they use that they can do it. 00:25:12.756 --> 00:25:15.253 You don't have to have a third party do this. 00:25:15.253 --> 00:25:17.970 You bring the tool into your own facility. 00:25:17.970 --> 00:25:19.433 Your own people are doing this. 00:25:19.433 --> 00:25:23.449 They're being able to use and access this data in the same way that you know. 00:25:23.449 --> 00:25:27.445 You get them all up and running, and then the intent is that they got this on their own. 00:25:28.027 --> 00:25:29.067 Yeah, yeah, and that's. 00:25:29.067 --> 00:25:30.148 That's the thing, right. 00:25:30.148 --> 00:25:33.513 There's no reason, once you have it, do it on your own. 00:25:33.513 --> 00:25:36.938 If you want to have somebody else, come in and do it, great, it's a great tool. 00:25:36.938 --> 00:25:38.819 You know it's a force multiplier. 00:25:38.819 --> 00:25:41.586 It's not. 00:25:41.586 --> 00:25:42.469 You're not looking to replace anybody. 00:25:42.469 --> 00:25:51.396 We're just looking to help make things run easier and faster, because really that's what it comes down to how fast can you do something with the least amount of effort within that affects the organization? 00:25:51.785 --> 00:26:26.704 Always, always no-transcript. 00:26:27.325 --> 00:26:30.755 What that means really is that you don't come out and brush it right. 00:26:30.755 --> 00:26:41.615 It's not like you have to clean it up because it got dirty, but it's that the data you have is applying to what you need, right, so it's not a lot of fodder in it. 00:26:41.615 --> 00:26:50.220 The way we go about it is when you're looking at everything, trying to dig through these haystacks right and figure this out. 00:26:50.220 --> 00:26:54.289 So if you start looking at trends, you start looking at patterns. 00:26:54.289 --> 00:27:05.593 We get interesting ability of running these intelligent queries, and part of the capability is not just well, we look at the file system information, or we look at the extended metadata, which is powerful to start with. 00:27:05.593 --> 00:27:07.686 But then let's even go deeper. 00:27:07.686 --> 00:27:10.614 You start looking at you know, introspecting a file. 00:27:10.614 --> 00:27:17.817 Okay, I'm going to open it, see what's here, start looking at the trends and we start identifying okay, what's relevant. 00:27:17.817 --> 00:27:21.518 It starts to become pretty interesting to identify. 00:27:21.578 --> 00:27:31.949 Then we're looking for everything that has this attribute with it and it could be an owner, who created it, what department it came from, what was the time frame, what was the project. 00:27:32.430 --> 00:27:37.980 All of that becomes now relevant for how you can classify the data. 00:27:37.980 --> 00:27:39.269 And that really is it. 00:27:39.269 --> 00:27:43.691 How do you classify right and then, with that, apply governance. 00:27:43.691 --> 00:27:44.673 How are we going to manage it? 00:27:44.673 --> 00:27:45.757 Is this in band? 00:27:45.757 --> 00:27:50.096 Is this going to be part of our project or is it going to be, you know, out of range? 00:27:50.096 --> 00:27:55.017 It shouldn't be part of the project because it includes personal information. 00:27:55.017 --> 00:28:08.634 You know, I know, we don't want anything that's going to include people's social security numbers or their credit card information or anything about their HR information that sometimes can get included unknowingly. 00:28:08.634 --> 00:28:33.554 So if you start having the ability of using filters, like we're doing, you can then get very granular on how you start to create this data and, as we're identifying, it we're grouping it using the custom metadata fields to then group it into a curated set and then take that curated set and even fine tune it down further. 00:28:35.106 --> 00:28:36.432 And it's all being done easy. 00:28:36.432 --> 00:28:47.414 It's using the power of metadata and creating unique schemas and fields that we can create these granular, curated data sets and do it very easily. 00:28:47.414 --> 00:29:06.970 Again, it only takes one or two people to be able to go through this and start to fine tune it and then, when you're applying it, you're applying it not just across each file at a time, you're applying it now across hundreds to thousands of files at a time to create that set. 00:29:06.970 --> 00:29:10.040 So that's where it really becomes of the quality. 00:29:10.040 --> 00:29:11.826 Is that you're getting something. 00:29:11.826 --> 00:29:21.192 That is part of answering the question or solving the problem that you've identified to create the best possible outcome. 00:29:21.865 --> 00:29:27.253 Once you've got your final data curation, identify that we've got quality, we've got everything we need. 00:29:27.253 --> 00:29:32.396 Now you can use the workflow engine of CrystalLocks to say, okay, what are we going to do? 00:29:32.396 --> 00:29:44.887 We're going to start moving the files, we're going to copy them from where they are and then copy them to somewhere else, and Dell will go in and we have what we call an out chest, which is really kind of a checkout. 00:29:44.887 --> 00:29:48.297 It's a copy out and you can copy out to any system. 00:29:48.297 --> 00:30:01.053 If you're going up to Snowflake Databricks which is a popular use other GPU providers, other cloud providers you have something you're going to move into a parallel file system and run through GPU locally. 00:30:01.053 --> 00:30:11.613 It just gives you all the flexibility you need to now take that within weeks, this curated data set that now you're applying to your solution of choice. 00:30:11.613 --> 00:30:16.734 So it just really streamlines that whole project from start to finish. 00:30:17.757 --> 00:30:18.117 Got it. 00:30:18.117 -->

00:30:28.170 So you get to scan and identify all the data extremely quickly at any repository whatsoever, and then you're using that metadata to. 00:30:28.170 --> 00:30:34.279 You can change your metadata, you can categorize it, you can protect data, certain data, critical data. 00:30:34.279 --> 00:30:54.727 You can utilize that as your mechanism to do anything that you need, and what that does is it gives you that power then to make really good decisions with your data and where you want to move it, what you want to do with it, how do you want to categorize it, how do you want to organize that for AI or for any other tool whatsoever, for any kind of product that you need. 00:30:54.727 --> 00:30:55.830 So it's so. 00:30:55.830 --> 00:30:59.146 It really gives you the ability to then automate things too. 00:30:59.146 --> 00:31:12.393 So if you need to constantly move data that's coming in from one source or from one site to another site, you can do that continually, and it's really a lot less manpower than the way organizations are doing this today. 00:31:12.955 --> 00:31:16.288 Correct, and that even includes like for really sensitive information. 00:31:16.568 --> 00:31:33.015 right, we've got a bunch of different sensitive data and files creating a data room, right, which is kind of we're going to isolate this so that we want to run through an AI project but we don't want to run and create any impact, right? 00:31:33.015 --> 00:31:34.411 Okay, so wonderful. 00:31:34.411 --> 00:31:47.390 You can actually use the virtualization power of Bristol Onyx to now even export that as a file system, so you can create a whole virtual data room within your place and not have to actually build or separate it. 00:31:47.390 --> 00:31:51.817 So it just gives you a lot of capabilities to do things a little bit differently. 00:31:52.685 --> 00:32:16.818 You said you should think differently right there, but that's one thing I really want people to take away from our conversation today is why do organizations need to rethink their AI approach Right and so for companies looking to adopt AI or starting to dive into this, what's your biggest piece of advice for either these leaders or even just managers within an organization that are looking to use AI? 00:32:17.339 --> 00:32:22.063 The biggest thing is make the decisions that are aligned to what you're trying to accomplish. 00:32:22.063 --> 00:32:41.038 Align to the business, because everything I don't care what it is, every piece of technology takes a certain approach and that certain approach is the way they are approaching a problem that may or may not align with the way the company is operating, and this includes us too. 00:32:41.038 --> 00:32:47.576 The way we're approaching is very, very specific and we see that it aligns for a lot of companies, a lot of customers. 00:32:47.576 --> 00:32:50.994 That doesn't mean it's the de facto standard for everyone. 00:32:50.994 --> 00:32:59.028 So what I do really quickly is, when we first talk to somebody is I'm understanding, okay, what are your top three challenges? 00:32:59.028 --> 00:33:00.830 How can we help and resolve? 00:33:00.830 --> 00:33:11.961 And really quickly I just can tell it's either we're going to align very well or it's going to intersect, and at that point, if it's intersection, it doesn't make sense. 00:33:11.961 --> 00:33:32.173 It doesn't matter, and I see this over and over again. 00:33:32.173 --> 00:33:35.258 Trying to take a round peg into a square hole. 00:33:35.258 --> 00:33:39.506 Right, it takes a lot of effort, you break things along the way. 00:33:39.506 --> 00:33:41.190 It might not be the best way. 00:33:41.190 --> 00:33:53.631 So taking a step back and saying let's approach this a little bit differently, and that's really the great thing is ensure alignment, and that includes not just with the culture and the operation, but also the people. 00:33:53.631 --> 00:33:57.147 And then how do you remove unnecessary steps? 00:33:57.790 --> 00:34:06.971 So one of the big things that we look at, and part of this whole idea of the round of virtualization, is why do we need to come and copy everything out first? 00:34:06.971 --> 00:34:11.226 Why don't we figure out what we can do, what's needed? 00:34:11.226 --> 00:34:16.434 How do we start to create the data set and curate it and classify it before we start the move? 00:34:16.434 --> 00:34:18.940 I mean, that saves so much time. 00:34:18.940 --> 00:34:32.391 It goes from this idea about data lakes right, moving everything up and creating a lake so that you have the ability to do the same thing, so we can get away from silos, get away from disparate systems or put it all together in one lake. 00:34:32.391 --> 00:34:51.929 But the problem is, when you do that, your lake starts to become swamp, and so what we decided is we have the ability to actually virtualize it and create a virtual data lake without ever having to move, and then get very clear on what this is, using the tools that we have, and then start to then move that up so it becomes. 00:34:51.949 --> 00:35:11.213 You know, it's like crystal clear waters at that point, and we found that just by doing it this way, companies actually start to move faster, start to leverage AI, leverage all the different projects and the things that they're wanting to do, and it's without even they don't even have to now outsource their data to third parties anymore, right? 00:35:11.213 --> 00:35:26.576 So if you're really sensitive and you're worried about creating that these types of processes to move to third-party and clouds and what have you there's always a potential for security breach or that intellectual property or trade city can get out. 00:35:26.576 --> 00:35:28.833 So this keeps it within the four walls. 00:35:28.833 --> 00:35:36.677 So you have a lot of power and control on where you and how you are implementing these third-party. 00:35:36.677 --> 00:35:38.088 And I'm not saying don't use

them. 00:35:38.088 --> 00:35:45.315 I'm just saying let's use them and use all of this better to really get to the outcome that everyone's wanting to get to. 00:35:46.246 --> 00:35:46.907 Yep, yep. 00:35:46.907 --> 00:35:49.094 Use each for what their strong suit is. 00:35:49.394 --> 00:35:49.795 Exactly. 00:35:50.005 --> 00:35:51.692 And make sure you're focused on that. 00:35:51.692 --> 00:35:59.655 So you know we spend a lot of time on the front side of this right Data cleanup and all of that getting ready to use AI. 00:35:59.655 --> 00:36:30.335 And you know, let's talk a little bit here, as we kind of come to the close of this discussion, about the final step, right, closing the AI project cycle loop and I think you know you've talked about AI project cycle four and start the modeling step Now that Crystallonics is involved, and within a matter of weeks we've got data readiness is completed, the data classification steps have been completed, a quality data set has been curated, or data sets and now AI training can begin. 00:36:30.335 --> 00:36:32.826 So tell me what's next. 00:36:33.768 --> 00:36:35.914 Well, before we go to what's next. 00:36:35.914 --> 00:36:42.193 I know we kind of touched on this, but I would say don't assume AI will solve all of your problems. 00:36:42.735 --> 00:36:42.856 Okay. 00:36:43.496 --> 00:36:45.713 Right, you know, if you think it's a panacea, it's not. 00:36:45.713 --> 00:37:03.155 If your data is not ready, it's not quality, it's not set up, ready to go, and you're just focusing on the AI tools and the cool shiny things, it's going to fail, right, Because it really has to start with smarter data management, and that's really kind of what we were just talking about. 00:37:03.155 --> 00:37:11.132 And so the faster you can efficiently clean up, classify, curate, the better your AI outcome and the better your insights will be. 00:37:11.132 --> 00:37:13.849 And there's a consultant. 00:37:13.849 --> 00:37:15.155 Her name's Leah Pinzer. 00:37:15.155 --> 00:37:18.592 I came across her about a month ago. 00:37:18.592 --> 00:37:23.994 She's an AI advisor, she's former CDP of Microsoft and she had a great quote. 00:37:24.155 --> 00:37:24.597 I loved it. 00:37:24.597 --> 00:37:29.030 She says that AI isn't a solution, it's a force multiplier. 00:37:29.030 --> 00:37:32.795 And you look at what's going on and it really is. 00:37:32.795 --> 00:37:39.286 It's just amazing about what are the things that, if you do it right, the outcome you can create. 00:37:39.286 --> 00:37:47.246 And she goes on to say if your data goals and processes are broken, ai will amplify that chaos. 00:37:47.947 --> 00:38:29.458 And her number one item on her AI project checklist that she preaches is data readiness Make sure your data is clean, make sure it's structured, make sure it's accessible, and so that's the things that we started to focus on originally, and just we cut our teeth on it for high-performance computing and we're working in some

of the most challenging environments in the world, where we're going through and scanning hundreds of petabytes to even exabytes in a single pass and then being able to start taking action and copying and migrating and doing all the cool things with it, but doing it at a point where we're doing it at scale. 00:38:29.458 --> 00:38:44.288 And so I think, as you're going through and starting now to apply this from an enterprise standpoint, there's a lot of things that we've learned over the last number of years that are perfect lessons learned for creating a successful AI project. 00:38:44.849 --> 00:38:51.289 You know, now you're speaking my speak right, we're talking about business processes first, and then tools follow, right? 00:38:51.289 --> 00:39:08.733 Yes, make sure that you've got good ways of working, good best practices, all of that good stuff, and then you want to focus, like we've talked about, on your pain points, your problem statements, and then make sure that we really understand those well before we go and expect any kind of a magic wand to do anything for us. 00:39:08.733 --> 00:39:14.231 So we just really need to treat it like a true project and really scope it out properly and have clear expectations. 00:39:15.153 --> 00:39:17.231 Absolutely, completely, 100% agree. 00:39:17.784 --> 00:39:18.186 I love it. 00:39:18.186 --> 00:39:18.867 Tony. 00:39:18.867 --> 00:39:20.773 We've talked about a ton of things here today. 00:39:20.773 --> 00:39:22.978 We've had a lot of good takeaways. 00:39:22.978 --> 00:39:34.130 First, I'd love you know if people are more interested in learning about Crystallonics or having a conversation with you about anything that we've talked about today, or just the capabilities of your tool and your solution. 00:39:34.130 --> 00:39:35.936 Where can listeners learn more about Crystallonics? 00:39:36.264 --> 00:39:36.626 Absolutely. 00:39:36.626 --> 00:39:37.106 Thanks, brandy. 00:39:37.106 --> 00:39:40.235 They can invite them to come and visit our website. 00:39:40.235 --> 00:39:49.172 It's crystallonicscom C-R-Y-S-T-A-L-O-N-Y-Xcom, or send an email. 00:39:49.172 --> 00:39:56.579 Send it to info at crystallonicscom, again, c-r-y-s-t-a-l-o-n-y-x. 00:39:56.579 --> 00:39:58.380 That comes right to my email. 00:39:58.380 --> 00:40:05.585 I'll see it Respond back. 00:40:05.585 --> 00:40:06.467 Let us know what we can do to help. 00:40:06.467 --> 00:40:07.329 Any questions, anything we can do. 00:40:07.329 --> 00:40:14.876 We're here to help and I just love talking with different people, different organizations, because every organization has a chance not only just to share what we're doing, but also learn more. 00:40:17.324 --> 00:40:18.208 Yeah, I highly encourage everyone to do so. 00:40:18.208 --> 00:40:25.516 I know Tony's been working on some rapid start programs as well to help organizations get a taste test of the benefits of what Crystallonics can provide. 00:40:25.516 --> 00:40:31.431 And you can do that very quickly and get a little trial, a little pilot run for you before you dive deep. 00:40:31.431 --> 00:40:33.670 Yeah, thank you, that's a great way to start. 00:40:33.885 --> 00:40:45.378 I mean it's low risk, right, because we kept thinking we have this great thing that we spent just years working on that we think we find really fine-tuned it and encouraged and wanted to open this up. 00:40:45.619 --> 00:40:57.237 So you know, whenever you start any type of new technology right and implementation goes through an evaluation, we're not going to just risk everything we're doing, put something in and say, well, okay, well, I hope it works. 00:40:59.206 --> 00:41:02.755 So, thinking through this, I thought, well, let's come up with a different approach on this. 00:41:02.755 --> 00:41:04.893 And so let's do a rapid start. 00:41:04.893 --> 00:41:39.675 And so, for a very nominal fee, you can go have three months, download the software, install it on your system in your environment either whether it's a VM, on a bare metal server, even on a cloud compute, in whatever works for you and then you'll be up and running within two to three hours and then within another hour, two hours after that, you've now scanned probably a couple million different files within your system and can actually start taking steps through this. 00:41:39.675 --> 00:41:48.414 So within half a day you're already on your way to start getting ready for data readiness and just start that process of okay, how do we apply it? 00:41:48.414 --> 00:41:52.855 How do we take this and start applying our model and even start playing with it. 00:41:52.855 --> 00:41:55.992 I encourage everyone try it out. 00:41:55.992 --> 00:41:58.653 It's a rapid start program you can see on the website. 00:41:58.653 --> 00:42:01.534 Send me an email at info at crystallonicscom. 00:42:01.534 --> 00:42:06.172 I'll send you a link to get started and then we can talk further. 00:42:06.773 --> 00:42:07.074 I love. 00:42:07.074 --> 00:42:14.373 It All right, tony, thank you so much, appreciate the time today as always, and we look forward to talking to you again here back on the podcast. 00:42:15.094 --> 00:42:15.577 Great Brandy. 00:42:23.824 --> 00:42:24.266 Thank you so much. 00:42:24.266 --> 00:42:24.987 Great to speak with you. 00:42:24.987 --> 00:42:25.347 Take care Bye. 00:42:25.347 --> 00:42:26.190 Thank you for tuning in today. 00:42:26.269 --> 00:42:32.945 Don't forget to subscribe and review the show and for more information on IPX, visit IPXHQcom.