TRANSKRIPT 🔗
1
00:00:02,584 --> 00:00:10,977
Today we have a special guest, Brad Magnetta, founder and CEO of Modly, a pioneering
company in the field of machine learning and artificial intelligence.
2
00:00:10,977 --> 00:00:20,731
With a PhD from Yale University, Brad's background as a research scientist in machine
learning has equipped him with the skills to innovate and disrupt traditional processes in
3
00:00:20,731 --> 00:00:21,862
ML development.
4
00:00:21,862 --> 00:00:27,994
His company, Modly, focuses on knowledge preservation to make ML development more
efficient and collaborative.
5
00:00:27,994 --> 00:00:30,165
Brad, it's great to have you on the podcast.
6
00:00:31,326 --> 00:00:33,788
Thanks so much for having me here, Jeff.
7
00:00:34,272 --> 00:00:36,536
All right, well, let's kind of dive into your journey.
8
00:00:36,536 --> 00:00:45,821
I guess tell me a little bit about, you know, your journey from academia to
entrepreneurship and kind of what motivated you to take the leap into entrepreneurship.
9
00:00:47,009 --> 00:00:58,228
Yeah, I mean, I've always been passionate about theorizing and testing ideas that really
drove me through grad school, studying physics, machine learning, and later as a research
10
00:00:58,228 --> 00:00:59,879
scientist in the industry.
11
00:00:59,920 --> 00:01:07,845
When I transitioned to industry, I found that the immediacy of seeing my ideas have a real
world impact was incredibly fulfilling.
12
00:01:07,866 --> 00:01:17,133
Unlike academic research, which often focuses on long-term goals, industry research
allowed me to work on innovative ideas with tangible, short-term
13
00:01:17,142 --> 00:01:19,014
outcomes, which really enjoyed.
14
00:01:19,295 --> 00:01:29,417
The balance of the experience of balancing priorities, cost, timelines, efficacy, and
applied research, it all made me realize that entrepreneurship is actually the perfect
15
00:01:29,417 --> 00:01:33,232
path for me to bring my ideas and deep tech to life.
16
00:01:33,232 --> 00:01:34,653
So, yeah.
17
00:01:34,827 --> 00:01:35,938
Yeah, definitely.
18
00:01:35,938 --> 00:01:43,795
And so tell me about some of like the key turning points of kind of the realization that
kind of led to the foundation of Modley.
19
00:01:45,707 --> 00:01:57,726
One of the big turning points for me was realizing that the traditional way of developing
ML and deep learning was pretty inefficient and not very collaborative.
20
00:01:58,787 --> 00:02:01,109
That's what really pushed me to start ModLin.
21
00:02:01,109 --> 00:02:10,076
I wanted to create a set of tools that could streamline the processes and make it easier
for teams to work together and really to build on each other's successes instead of
22
00:02:10,076 --> 00:02:11,697
working in silos.
23
00:02:13,741 --> 00:02:14,623
Got it, okay.
24
00:02:14,623 --> 00:02:23,624
And tell me a little bit about how, I guess, your background as a research scientist
influenced your approach to building a company focused on machine learning and AI.
25
00:02:25,581 --> 00:02:31,245
think my research background is definitely shaped how I approach building a company.
26
00:02:31,245 --> 00:02:34,527
In research, you learn to be meticulous and methodical.
27
00:02:34,707 --> 00:02:42,652
And that's been crucial in making sure our innovations are not only cutting edge, but also
practical, scalable, and a
28
00:02:44,331 --> 00:02:45,743
Yeah, definitely.
29
00:02:45,743 --> 00:02:50,731
And I know we talked a lot about Techstars previously when we were preparing for this
podcast.
30
00:02:50,731 --> 00:02:56,560
Tell me a little bit about that experience and really how it contributed to your personal
growth, your company's growth.
31
00:02:57,975 --> 00:03:02,757
The Techstars was such a great experience for me personally, but also a game changer for
Mavli.
32
00:03:02,757 --> 00:03:05,409
It wasn't just about the mentorship.
33
00:03:05,409 --> 00:03:12,072
It was about being part of a community that really challenged us to refine our vision,
push our limits.
34
00:03:12,372 --> 00:03:17,444
The connections we made and the feedback we received helped us scale much faster than we
could have on our own.
35
00:03:17,444 --> 00:03:23,187
So it was such a great experience and really feel fortunate to be a part of that Boston
cohort.
36
00:03:23,328 --> 00:03:24,321
Yeah, definitely.
37
00:03:24,321 --> 00:03:30,451
And are you still in touch with people from Techstars and is it still kind of integrated
into your life?
38
00:03:31,913 --> 00:03:38,703
yeah, mean, everyone in the group of founders is so busy, but it's great to keep tabs from
time to time.
39
00:03:38,703 --> 00:03:41,807
And it was a really great cohort to be a part of.
40
00:03:41,948 --> 00:03:48,657
Definitely all of us really bonded and had a, and keep in touch from time to time.
41
00:03:48,938 --> 00:03:49,848
Nice.
42
00:03:49,969 --> 00:03:50,539
Alrighty.
43
00:03:50,539 --> 00:03:55,173
Well, let's talk a little bit about some of your, I guess, business insights.
44
00:03:55,173 --> 00:03:59,246
You mentioned kind of knowledge preservation as a core aspect of Modly.
45
00:03:59,246 --> 00:04:04,230
Can you explain how this concept is transforming ML development?
46
00:04:05,889 --> 00:04:09,291
Yeah, we're really focused on what we call mileage preservation.
47
00:04:09,372 --> 00:04:14,596
It's all about capturing and reusing what's already done to avoid repeating the same work.
48
00:04:14,596 --> 00:04:20,000
This approach has been a game changer for ML development, making projects faster, more
collaborative.
49
00:04:20,000 --> 00:04:22,963
It's like giving teams a head start with every new project.
50
00:04:22,963 --> 00:04:30,529
So you never lose work, lose a place in your work, and you can always have someone take
off where you left off.
51
00:04:30,529 --> 00:04:31,429
That kind of
52
00:04:32,246 --> 00:04:33,497
Got it, okay.
53
Load full transcript
00:04:33,879 --> 00:04:41,891
And in, I guess, the early stages of your company, what were some of the biggest
challenges that you faced and tell me a little bit about how you overcame them?
54
00:04:43,711 --> 00:04:48,263
I think in research, you're often focused on pursuing a big vision.
55
00:04:48,263 --> 00:04:51,544
What we're pursuing is called meta-learning.
56
00:04:51,624 --> 00:05:01,949
But one of the biggest challenges of building a company around an innovative idea is
breaking that vision down into actionable steps that also deliver a business impact along
57
00:05:01,949 --> 00:05:02,849
the way.
58
00:05:03,249 --> 00:05:09,912
This becomes even more challenging when you're working with small teams, limited time, and
have a long list of things to accomplish.
59
00:05:09,912 --> 00:05:11,773
So we definitely
60
00:05:11,797 --> 00:05:18,237
worked through what was the right way to strategize building towards our vision, figured
it out.
61
00:05:18,237 --> 00:05:23,946
that's definitely something that's difficult when you're building a company around
research.
62
00:05:23,946 --> 00:05:27,871
So that was one of our early challenges that was.
63
00:05:28,724 --> 00:05:30,147
Yeah, that makes sense.
64
00:05:30,147 --> 00:05:37,578
And how have you navigated, I guess, the balance between maintaining your core vision and
the pressures of monetizing your innovations?
65
00:05:38,967 --> 00:05:40,898
Yeah, it's always a tough one.
66
00:05:40,898 --> 00:05:47,522
Balancing our vision with the need to generate revenue has certainly been a juggling act
from time to time.
67
00:05:47,522 --> 00:05:49,763
I think every founder feels that.
68
00:05:50,204 --> 00:05:54,285
However, we believe there's always a way to monetize innovative technology.
69
00:05:55,406 --> 00:06:02,030
And we focused on developing both community and enterprise tools from the same core
technology.
70
00:06:02,030 --> 00:06:06,172
A great example of that is a technology virus called Modly Eval.
71
00:06:06,989 --> 00:06:12,991
And this really allows us to grow a strong user base while also driving revenue at this
stage of our growth.
72
00:06:12,991 --> 00:06:26,007
So it's definitely a balancing act, but there's always a way to figure out how to build
your business to deliver revenue while not compromising the technology to build towards
73
00:06:26,007 --> 00:06:26,867
your vision.
74
00:06:28,153 --> 00:06:33,606
And I know we talked a little bit about Modly Eval before and the impact that it's had on
your business.
75
00:06:33,606 --> 00:06:36,298
Tell me a little bit about more about Modly Eval.
76
00:06:37,527 --> 00:06:53,608
Yeah, know, Modly Evel is really a tool that was born out of a pain that I felt as a
founder, even with my expertise in ML, trying to find and hire really qualified, great ML
77
00:06:53,608 --> 00:06:54,729
employees.
78
00:06:54,729 --> 00:07:06,177
It was really hard not having an HR department or a head of talent, you know, to help me
through that process.
79
00:07:06,443 --> 00:07:12,696
I found myself, the first time I did it, not wasting a lot of time, but spending a lot of
time through the interview process.
80
00:07:12,696 --> 00:07:17,778
So the second time through, I knew that I needed a more systematic way.
81
00:07:17,778 --> 00:07:29,342
And so what I did is I basically hacked together a product around our core technology that
gives out an assessment, preserves the knowledge of all of the individuals that take the
82
00:07:29,342 --> 00:07:36,039
assessment, compares their work against each other so that I can not only provide
feedback,
83
00:07:36,039 --> 00:07:41,611
systematically, but also can easily tell how well they stand against each other.
84
00:07:42,191 --> 00:07:44,972
Just giving them an exercise.
85
00:07:45,272 --> 00:07:51,194
So this has really been a game changer for the way that we hire and find really quality
talent here at Mobley.
86
00:07:51,194 --> 00:07:59,807
We were really excited to find a solution for our pain and then start to offer that to
other organizations like ourselves and more.
87
00:07:59,914 --> 00:08:01,174
Yeah, I love that.
88
00:08:01,174 --> 00:08:05,654
And I love how much I love just everything about it, saving time.
89
00:08:06,034 --> 00:08:09,614
We've been creating a lot of custom assessments for people.
90
00:08:09,614 --> 00:08:18,854
So I think you said before, you have like hundreds of applicants coming in, you send them
the assessment and then you have bell curve and you can see kind of maybe who's in the top
91
00:08:18,854 --> 00:08:19,954
five to 10%.
92
00:08:19,954 --> 00:08:23,202
And then you can proceed with interviews with that.
93
00:08:23,344 --> 00:08:24,976
subset of the group before.
94
00:08:24,976 --> 00:08:27,969
So you're saving a lot of time conserving energy.
95
00:08:27,969 --> 00:08:31,752
Is that kind of the right way in terms of how it works for you?
96
00:08:32,813 --> 00:08:43,133
100%, I think it's almost creating what we call like a quantitative filtering of
candidates in the ML space, which is a really hard thing because at the moment all you
97
00:08:43,133 --> 00:08:51,213
have is your resume, a bunch of resumes to look through, maybe some filters on levels of
experience that the candidates fill out.
98
00:08:51,213 --> 00:09:00,449
But there's still really no way to tell how well each of your individual candidates
quantitative skills are in ML until
99
00:09:00,631 --> 00:09:03,813
You already filter them by school or by location.
100
00:09:03,813 --> 00:09:06,364
And so really what this does, it kind of flips it.
101
00:09:06,364 --> 00:09:12,928
Instead of doing technical evaluation at the end with a smaller pool of candidates, we put
it to the front.
102
00:09:13,088 --> 00:09:19,371
And what that does is it gives everyone a fair shot, regardless of their background,
allows them to stand out.
103
00:09:20,312 --> 00:09:25,915
It gives them feedback so that it can actually learn from the experience, see where they
fit on the bell curve.
104
00:09:25,915 --> 00:09:28,577
And then we have all sorts of great resources.
105
00:09:29,663 --> 00:09:34,497
in our Modly documentation website, exercises, ways to get better.
106
00:09:34,497 --> 00:09:36,338
So really, we're kind of flipping out on its head.
107
00:09:36,338 --> 00:09:39,901
We're allowing you to filter people quantitatively to start.
108
00:09:39,901 --> 00:09:41,743
And I think that it's great for the candidate.
109
00:09:41,743 --> 00:09:43,685
It's great for the employer.
110
00:09:43,685 --> 00:09:48,408
And it helps people find people and organizations find each other.
111
00:09:49,564 --> 00:09:50,805
Yeah, no, I love that.
112
00:09:50,805 --> 00:09:57,818
And as a recruitment agency, you know, we do that by, like you said, by universities or by
company names.
113
00:09:57,818 --> 00:10:02,950
You know, we know that the bar at this company is high, so we're going to target these
company names only.
114
00:10:02,950 --> 00:10:10,124
And if we don't recognize the company, if we don't know anything about their technical
bar, then we're not going to move forward with the candidate because we don't know that
115
00:10:10,124 --> 00:10:10,434
team.
116
00:10:10,434 --> 00:10:10,704
Right.
117
00:10:10,704 --> 00:10:15,486
And so I love that this really opens that up and gives people the opportunity.
118
00:10:15,570 --> 00:10:20,715
to get in that traditionally haven't been able to get into some of these companies.
119
00:10:22,238 --> 00:10:32,189
Well, I know, guess given the current economic climate, tell me a little bit about how you
successfully secured funding and kind of what advice you would give to others looking to
120
00:10:32,189 --> 00:10:33,220
do the same.
121
00:10:35,095 --> 00:10:45,368
Yeah, it's really tough right now, but I don't think that is a reason to not try for
anyone that's interested in trying.
122
00:10:46,428 --> 00:10:59,592
But particularly in today's economic climate, you really have to demonstrate that, and we
did, that we not only had a strong product, but also a clear and realistic path to growth.
123
00:11:00,652 --> 00:11:03,833
And my advice to others is to really focus on solving real
124
00:11:03,927 --> 00:11:16,248
tangible problems, especially those you have a direct connection to and a problem that you
understand deeply, because that matters a lot when you're talking with investors.
125
00:11:17,190 --> 00:11:29,171
It's crucial to clearly articulate the value your solution brings and why you're uniquely
positioned to deliver it, especially today in such a tough climate.
126
00:11:29,352 --> 00:11:30,883
Yeah, definitely.
127
00:11:31,403 --> 00:11:34,825
Well, let's talk a little bit about some of the technical areas.
128
00:11:34,825 --> 00:11:42,590
What are some of the technical challenges that you've encountered in developing an AI for
AI system and how have you addressed them?
129
00:11:44,011 --> 00:11:53,898
Yeah, I think building our core technology to support meta-learning, which is essentially
an AI for AI in a scalable way, was a significant challenge.
130
00:11:53,898 --> 00:11:59,992
It required us to initially limit our focus to a few specific machine learning use cases.
131
00:12:00,452 --> 00:12:07,137
This focus allowed us to hone our approach, and secure and ensure a strong foundation for
technology, though.
132
00:12:07,297 --> 00:12:13,421
However, we're continuously innovating and improving, and I'm incredibly proud of what
we've accomplished so far.
133
00:12:13,581 --> 00:12:20,746
Sooner, technology will be equipped to handle an even broader range of use cases,
including those involving large language models.
134
00:12:20,746 --> 00:12:24,358
So you really have built a very strong foundation.
135
00:12:24,358 --> 00:12:34,755
But you have to be kind of turn that technical challenge into building and selecting the
right foundation to build, because you can't do everything at once.
136
00:12:34,755 --> 00:12:38,527
So it took a lot of strategy and a lot of great work from my team.
137
00:12:38,846 --> 00:12:40,028
Yeah, definitely.
138
00:12:40,028 --> 00:12:43,474
You talked a little bit about knowledge preservation before, I guess.
139
00:12:43,474 --> 00:12:51,065
Can you kind of break down the process of knowledge preservation and how it significantly
reduces repetitive efforts in ML?
140
00:12:52,769 --> 00:12:57,540
Yeah, knowledge preservation at Modly is about more than just saving data.
141
00:12:57,540 --> 00:13:05,313
It's about capturing the strategies and insights that led to success in previous machine
learning development.
142
00:13:05,313 --> 00:13:10,654
This way teams can build on what's already been done instead of starting from scratch
every time.
143
00:13:10,894 --> 00:13:18,006
I like to think about knowledge preservation in contrast to existing ML ops, like
experiment documenting technology.
144
00:13:18,006 --> 00:13:21,017
We really set a standard for which
145
00:13:21,965 --> 00:13:32,732
for what you should and all ML developers should document about the experiments and the
models that they train so that their work is comparable to others.
146
00:13:32,732 --> 00:13:35,693
And you can only really do that through standardization.
147
00:13:36,534 --> 00:13:43,458
And as a result, we call what we're doing here at Maldives, MLOps, knowledge preservation.
148
00:13:43,458 --> 00:13:48,561
So it really does preserve those insights so they're comparable so that insights can be
derived.
149
00:13:49,274 --> 00:13:50,235
Got it.
150
00:13:50,959 --> 00:13:57,817
And how do you foresee the kind of integration of Modly's tools in everyday machine
learning projects across different sectors?
151
00:13:59,629 --> 00:14:10,825
So Mobly's core technology is a Python package that integrates seamlessly with popular
machine learning frameworks like PyTorch and MLflow, making it a valuable tool for ML
152
00:14:10,825 --> 00:14:11,976
development.
153
00:14:12,156 --> 00:14:21,341
While we currently excel in supporting industries focused on image-based machine learning
solutions, we're actively expanding to support additional use cases, including tabular,
154
00:14:21,341 --> 00:14:23,062
time series, text, and more.
155
00:14:23,062 --> 00:14:28,637
And you'll actually see a push, the latest version of our package that we're pushing.
156
00:14:28,637 --> 00:14:30,538
very soon is going to show that.
157
00:14:30,798 --> 00:14:42,925
Our tools are designed to give junior developers really a stronger starting point by
leveraging insights from the broader community or senior developers within their company,
158
00:14:42,925 --> 00:14:57,393
which really enables them to accelerate their learning and productivity and contribute to
higher difficulty tasks sooner at early stages of their career.
159
00:14:58,822 --> 00:14:59,722
Got it.
160
00:14:59,722 --> 00:15:05,622
And what, I guess, what innovations do you see in your space that you expect to see kind
of by 2030?
161
00:15:05,622 --> 00:15:09,720
And how is Modly preparing for some of these advancements?
162
00:15:11,243 --> 00:15:24,620
Yeah, while generative AI has been an absolute groundbreaking advancement in the field of
ML, we're hearing from customers that using such heavy technology to tackle niche problems
163
00:15:24,680 --> 00:15:28,022
often isn't sustainable, scalable, or cost effective.
164
00:15:28,022 --> 00:15:40,863
There certainly are going to be uses of generative AI where you'll want to use chat GPT,
you'll want to use big heavy AI and ML, but not every problem
165
00:15:40,863 --> 00:15:44,255
needs a genius to solve, right?
166
00:15:44,716 --> 00:15:52,621
If you're looking at a series of images or paragraphs and you're making just a quick
decision, does it fit in this bucket or that bucket?
167
00:15:52,681 --> 00:15:57,605
It doesn't require the genius of a really complex AI.
168
00:15:57,605 --> 00:16:09,899
And what you'll find is a lot of ML pipelines through industry are really a patchwork
quilt of all these
169
00:16:09,899 --> 00:16:12,370
really difficult decisions with minor ones.
170
00:16:12,370 --> 00:16:26,676
And I really see that a major need in the future is distilling the knowledge and power of
these really complex AIs, these really impressive AIs, into smaller problems specific
171
00:16:26,676 --> 00:16:28,897
solutions that are ready for production.
172
00:16:28,897 --> 00:16:30,537
So why would you do that?
173
00:16:30,758 --> 00:16:32,378
It's cost effective.
174
00:16:32,499 --> 00:16:38,733
It's less time to deliver a prediction or solution to your customer.
175
00:16:38,733 --> 00:16:47,317
So there's a lot of advantages of still distilling that knowledge into what I've been
calling like deep learning equivalence, even though it all kind of
is deep learning.
176
00:16:49,039 --> 00:16:58,854
This really aligns perfectly with Mobly's mission as our platform is designed to
streamline and simplify this process, making it easier to create efficient targeted ML
177
00:16:58,854 --> 00:17:00,865
solutions customly.
178
00:17:01,025 --> 00:17:08,437
And you'll find over this next quarter and through the rest of this year, we're actually
going to be building and bridging our connection
179
00:17:08,437 --> 00:17:15,569
of our current package, which is more deep learning focused, to LLMs and sort of
agent-like use of them.
180
00:17:16,102 --> 00:17:17,062
Got it.
181
00:17:17,644 --> 00:17:21,048
Well, let's dive a little bit into your personal journey here.
182
00:17:21,048 --> 00:17:34,185
Tell me a little bit about, I guess, your leadership style, I guess, any kind of maybe how
your personal journey and your background has influenced your leadership style at
183
00:17:34,185 --> 00:17:35,106
Modeling.
184
00:17:37,089 --> 00:17:46,656
think my leadership style at Modly is really based on building trust between me and all
the people I work with.
185
00:17:46,656 --> 00:17:57,863
When you establish that trust, you can delegate tasks, you can give people creative
freedom, you can not have to do everything, right?
186
00:17:57,863 --> 00:18:01,205
And as a founder, you can't do everything, right?
187
00:18:01,205 --> 00:18:06,719
So just constantly trying to communicate, making sure that my employees
188
00:18:07,057 --> 00:18:09,758
my coworkers and I are on the same page.
189
00:18:11,659 --> 00:18:16,901
And just once that trust is built, it's a lot easier to kind of build together.
190
00:18:16,901 --> 00:18:26,045
So that's a really like main theme of my leadership style and I guess me personally.
191
00:18:26,318 --> 00:18:30,109
Yeah, remind me how many years have you been in business?
192
00:18:31,411 --> 00:18:34,706
We've been in business about a year and a half now.
193
00:18:34,827 --> 00:18:39,878
So it's not been a long time, but it's certainly flown by.
194
00:18:39,878 --> 00:18:44,198
Yeah, yeah, and I'm interested, you know, founding a company and starting a company in
2023.
195
00:18:44,198 --> 00:18:45,248
Did you start it?
196
00:18:45,248 --> 00:18:49,238
Do you guys have an office in person or do you hire wherever the talent is?
197
00:18:49,238 --> 00:18:50,782
What does that look like for you?
198
00:18:51,627 --> 00:18:56,501
Yeah, we're remote and we hire wherever the talent is.
199
00:18:57,622 --> 00:18:59,564
I really enjoy remote work myself.
200
00:18:59,564 --> 00:19:04,308
And I know that there's a lot of really quality remote workers.
201
00:19:04,308 --> 00:19:09,692
And I actually think that our tool, Modly Eval, is great for that.
202
00:19:09,692 --> 00:19:19,400
Regardless of location, whether it's US-based, whether it's global-based talent, our tool
really helps compare people.
203
00:19:20,981 --> 00:19:30,538
not just within your city, people from all around the US and the world and against the top
and all talent.
204
00:19:30,538 --> 00:19:43,477
we are remote, we have some co-working space, but I think we've cleverly built some tools
that help us make sure that we're a really effective team in that way.
205
00:19:43,658 --> 00:19:44,199
definitely.
206
00:19:44,199 --> 00:19:48,441
Yeah, we run a fully remote company now too with I think we're in five different countries
now.
207
00:19:48,441 --> 00:19:53,984
Our workforce is distributed across five different countries and I love that.
208
00:19:53,984 --> 00:19:57,485
Just really kind of hiring where the best talent is.
209
00:19:58,426 --> 00:20:06,596
In terms of, I guess, your most significant, I guess, life lessons that you've learned
since starting your own company, you know, you've...
210
00:20:06,596 --> 00:20:09,124
I know you've been on a of a wild journey in a short time.
211
00:20:09,124 --> 00:20:13,647
guess what would you say some of the biggest lessons that you've learned so far are?
212
00:20:17,707 --> 00:20:24,999
Yeah, I think just to be really honest, it's really hard.
213
00:20:24,999 --> 00:20:26,819
It's scary at times.
214
00:20:27,440 --> 00:20:34,462
And I know that these are just like, you know, I think that is kind of a lesson because
you go into this thinking, I'll figure it out.
215
00:20:34,462 --> 00:20:36,242
You will, but it's hard.
216
00:20:36,242 --> 00:20:46,165
I think it's super important as a result of all those things to find the right people to
build with that work well with you and how
217
00:20:46,655 --> 00:20:47,966
you need to build.
218
00:20:47,966 --> 00:20:53,711
When you're founding a company, you need to find great people to help you do that.
219
00:20:53,711 --> 00:20:55,231
And you need to find...
220
00:20:58,071 --> 00:21:00,672
people that match the way that you want to do that, right?
221
00:21:00,672 --> 00:21:08,127
So I think, you know, that's something I learned through the past year and a half, still
learning every day.
222
00:21:08,127 --> 00:21:10,938
And I think that's maybe the other takeaway is like never stop learning.
223
00:21:10,938 --> 00:21:12,609
You just have to embrace it.
224
00:21:12,909 --> 00:21:15,371
But I really think it's well worth it.
225
00:21:15,371 --> 00:21:16,911
Very fulfilling.
226
00:21:18,212 --> 00:21:22,515
A few years after PhD, I really wanted to keep challenging myself.
227
00:21:22,515 --> 00:21:26,413
I knew that this was like a challenge that I've been wanting to stay.
228
00:21:26,413 --> 00:21:30,127
for myself for a very long time, even through grad school.
229
00:21:30,127 --> 00:21:32,540
I always felt like I wanted to start a company.
230
00:21:32,540 --> 00:21:34,382
And you get what you ask for.
231
00:21:34,382 --> 00:21:37,385
It's tough, but I really enjoyed it.
232
00:21:37,697 --> 00:21:38,888
Yeah, definitely.
233
00:21:38,888 --> 00:21:46,010
That's probably
my favorite part about entrepreneurship too, is like having direct control
over like never stop learning.
234
00:21:46,010 --> 00:21:51,903
You can actually adopt like cutting edge technology and you could constantly be changing.
235
00:21:51,903 --> 00:21:56,565
I felt so trapped before I was like, I can't leverage this technology.
236
00:21:56,565 --> 00:22:00,616
I can't make these decisions because I, it's not getting approved, right?
237
00:22:00,916 --> 00:22:02,177
I try to take it up the chain.
238
00:22:02,177 --> 00:22:02,937
It's not working.
239
00:22:02,937 --> 00:22:07,600
So it's really awesome to like really fully embrace that.
240
00:22:07,600 --> 00:22:11,097
embrace the latest and greatest technology, things like that.
241
00:22:11,097 --> 00:22:12,460
What has been...
242
00:22:12,460 --> 00:22:12,725
Yeah.
243
00:22:12,725 --> 00:22:16,946
If I could say, just to add to that, I totally agree.
244
00:22:16,946 --> 00:22:33,181
I think the amount of stuff, especially in this time with this wave of just exciting stuff
that's happening in AI, we are doing so much cool stuff internally and able to operate
245
00:22:33,181 --> 00:22:40,772
with a small team doing things in marketing, decision making with LMs.
246
00:22:42,593 --> 00:22:51,222
with all that power, it's such a cool time to be building in such a transformative time of
technology.
247
00:22:51,222 --> 00:22:53,244
So I totally agree with that.
248
00:22:53,244 --> 00:22:54,215
Like that's one thing.
249
00:22:54,215 --> 00:22:58,840
I feel like I've been able to really embrace this technology and these tools.
250
00:22:58,840 --> 00:23:02,193
I don't know if I could have done that had I not started a company a year and half
251
00:23:02,527 --> 00:23:09,713
Yeah, and working for a small company where you can be agile, you can be nimble, you can
really be at the cutting edge.
252
00:23:10,054 --> 00:23:15,889
And I think it's only going to get faster, you know, these changes and these cycles of
technology.
253
00:23:15,889 --> 00:23:23,425
Now that we are where we're at with AI and these tools, it's just going to...
254
00:23:24,146 --> 00:23:28,149
The speed at which things are coming at us, I feel like it's going to be greater than
ever.
255
00:23:29,719 --> 00:23:39,155
Let's see here, in terms of maybe you're just like biggest mentors and influences, you
know, throughout your career, what's kind of made the biggest impact on you so far in
256
00:23:39,155 --> 00:23:44,643
terms of like mentorship and external influences outside of the company on your personal
growth?
257
00:23:45,879 --> 00:24:03,356
Yeah, I'll say one of my biggest mentors, Vinvids Oslins, my PhD advisor at Yale, he
really, he was a great role model for someone who can think creatively, but also really
258
00:24:03,356 --> 00:24:04,216
quantitatively.
259
00:24:04,216 --> 00:24:09,178
So I mean, just really well-rounded person, one of the most intelligent people I've ever
met.
260
00:24:09,178 --> 00:24:11,359
So I feel like I just learned so much from him.
261
00:24:11,359 --> 00:24:15,821
And I feel like I've always been a very creative thinker.
262
00:24:16,051 --> 00:24:21,085
And he really helped me hone sort of that quantitative side of me and sharpen it.
263
00:24:21,205 --> 00:24:22,556
And so, yeah, I don't know.
264
00:24:22,556 --> 00:24:39,981
I think a lot of what I learned through him and learned through grad school, like working
on my PhD, it's allowed me to apply those learnings to the business side of things, to ML,
265
00:24:39,981 --> 00:24:42,663
you know, not just science, because I started from a science background.
266
00:24:42,663 --> 00:24:44,905
So I really attribute.
267
00:24:45,013 --> 00:24:51,993
a lot of the way that I approach new things, new problems, like through that experience
with him.
268
00:24:51,993 --> 00:24:57,173
So he was a huge influence on me personally.
269
00:24:57,173 --> 00:24:58,005
Yeah.
270
00:24:58,005 --> 00:25:04,878
And I guess how do you continue to develop your skills and knowledge in this kind of
rapidly evolving field?
271
00:25:06,465 --> 00:25:22,998
I think once you get used to research, used to being at the edge of things, again, that
experience with him and other mentors, you just kind of start to figure out your style of
272
00:25:22,998 --> 00:25:23,979
approaching new things.
273
00:25:23,979 --> 00:25:28,582
So I feel like I have a blueprint for what works for me.
274
00:25:28,603 --> 00:25:32,266
Everyone has something, their own ways of doing things.
275
00:25:32,266 --> 00:25:36,269
But once you learn how to do that, it's easier to keep doing it.
276
00:25:36,269 --> 00:25:43,125
So I just kind of apply the same kind of principles of how I learn things.
277
00:25:43,125 --> 00:25:51,581
Once you learn how to learn, you can really learn anything if you have enough time and you
have enough patience.
278
00:25:53,183 --> 00:26:00,549
I think the easier or the longer or the more you try and do that, push your limits, the
easier it gets.
279
00:26:01,210 --> 00:26:03,962
And I've just been doing it for a while now.
280
00:26:03,962 --> 00:26:05,037
So it's.
281
00:26:05,037 --> 00:26:06,344
It's just second nature.
282
00:26:06,613 --> 00:26:07,764
Yeah, definitely.
283
00:26:07,764 --> 00:26:13,637
It's funny that you say once you learn how to learn and a lot of people never really learn
how to learn.
284
00:26:13,637 --> 00:26:25,935
They just, you know, they get taught in school and all of these things, but you never have
a class on teaching you how to learn and your your best way of how you learn.
285
00:26:25,935 --> 00:26:26,185
Right.
286
00:26:26,185 --> 00:26:31,578
I just saw a speech on with Jim Quick on that and about kind of learning how to learn and
that.
287
00:26:31,578 --> 00:26:33,199
So I love that you said that.
288
00:26:33,379 --> 00:26:43,099
Well, I guess kind of closing it out here and when you kind of reflect back on everything,
if you could go back to your 18 year old self and only take three lessons with you,
289
00:26:43,099 --> 00:26:49,309
whether it be philosophy, business knowledge or general advice, what would they be and
why?
290
00:26:50,711 --> 00:27:00,864
Yeah, I'd say all of this is kind of general advice, but don't let others define your
limits, I think is a really important one.
291
00:27:02,024 --> 00:27:08,896
think you really need to focus on proving to yourself what you're truly capable of, push
boundaries and explore your potential.
292
00:27:08,896 --> 00:27:17,948
And that's not like without unrealistic expectations.
293
00:27:17,948 --> 00:27:20,269
It's just know what you're capable of.
294
00:27:20,269 --> 00:27:22,970
capable of and no one can tell you that.
295
00:27:22,970 --> 00:27:25,470
You have to figure that out.
296
00:27:25,470 --> 00:27:28,271
So don't want to let others define your limits.
297
00:27:28,631 --> 00:27:30,712
Embrace failure as a stepping stone.
298
00:27:30,712 --> 00:27:32,582
It's really hard to get through those moments.
299
00:27:32,582 --> 00:27:37,234
I've even had moments that are tough, like through school, through career, things like
that.
300
00:27:37,234 --> 00:27:39,414
There's a lot of ups and downs for everyone.
301
00:27:39,914 --> 00:27:42,955
It's not easy, but each step back is really an opportunity.
302
00:27:42,955 --> 00:27:45,836
You just have to learn and you have to grow from it.
303
00:27:45,836 --> 00:27:47,356
So keep moving forward.
304
00:27:47,356 --> 00:27:49,677
No matter what, sometimes you have to stay flexible.
305
00:27:49,677 --> 00:27:50,787
and adaptable.
306
00:27:50,787 --> 00:27:54,497
So this is my third one to get through those moments.
307
00:27:55,777 --> 00:27:59,337
And I think that's a great thing too, just to know.
308
00:27:59,337 --> 00:28:06,917
Like when you're 18, thinking about what you're going to study in college, you can't
predict what your 25 year old self is going to be interested in.
309
00:28:06,917 --> 00:28:15,797
And then you see that a lot in young people now where they go to school for something,
they start a career in something they need to switch.
310
00:28:15,797 --> 00:28:16,617
Like you don't know.
311
00:28:16,617 --> 00:28:19,583
So you have to really stay flexible and adaptable.
312
00:28:19,765 --> 00:28:30,659
And if you really focus on learning things that generally interest you, but things that
also will make you adaptable, you'll have a really successful, happy career in life.
313
00:28:30,860 --> 00:28:32,371
So those three things.
314
00:28:32,448 --> 00:28:33,318
Yeah, I love that.
315
00:28:33,318 --> 00:28:38,428
And that's been coming up for me a lot, what you said about like not letting others create
your limits, right?
316
00:28:38,428 --> 00:28:45,928
Because I think the difference between us and all of these world record holders and these
things is they don't have a limiting belief around that, you know?
317
00:28:45,928 --> 00:28:52,188
There's people that can dive, free water dive for 15 minutes and it's because they don't
have a limiting belief, you know?
318
00:28:52,188 --> 00:28:54,298
They believe they can, so they do it, right?
319
00:28:54,298 --> 00:29:02,815
And it's amazing what happens when you break those beliefs or those limiting beliefs about
what you can accomplish.
320
00:29:03,203 --> 00:29:04,174
Well, I love this.
321
00:29:04,174 --> 00:29:05,546
This was a great conversation.
322
00:29:05,546 --> 00:29:10,652
know, really appreciate you sharing your thoughts and insights and experiences with us.
323
00:29:10,652 --> 00:29:13,836
It's been a pleasure having you on the Trailblazers podcast today.
324
00:29:14,945 --> 00:29:15,839
Yeah, thanks so much, Jeff.
325
00:29:15,839 --> 00:29:17,153
It was great being here.