Latest / AI for Founders with Ryan Estes / Future of privacy-first computer vision in security tech

Transcript

Future of privacy-first computer vision in security tech · AI for Founders with Ryan Estes

0:00
What's up founders? So I'm brighter today I got a
0:02
comment that said it was too dark in the YouTube comments and
0:05
now I'm feeling a little self-conscious.
0:06
Is it too bright? Is it too dark?
0:09
I don't know, We're working on the light.
0:10
I want to give a big shout out to the 25,000 folk who have
0:13
subscribed to AI for Founders. You can check that out.
0:16
AI for founders.co. We are one of the fastest
0:20
growing AI newsletters in the world today.
0:23
We've gone from zero to 25,000 since January and I'm super
0:27
stoked. I want to encourage the folks
0:29
that listen to the podcast. Check out the weekly newsletter.
0:32
It's, it's packed full of news of tools, summaries.
0:36
It's, it's amazing. We do a prompt of the week,
0:38
although I feel like I phoned in the prompt of the week this week
0:41
'cause it was just now do what I said the first time, because
0:46
sometimes that prompt works. Anyway, welcome back to AI for
0:50
Founders. Today we've got a powerhouse
0:53
guest, Galvin Wijaja, founder and CEO of Loretta dot IO.
0:58
Galvin took lessons from running restaurants in Singapore and
1:01
flipped them into building 1 of the world's most trusted privacy
1:04
first AI companies, working with clients like the US Department
1:08
of Homeland Security, TSA, and the Air Force.
1:11
In this episode, you'll learn how Loretta is redefining
1:14
computer vision, moving away from surveillance toward real
1:17
time understanding of human behavior.
1:19
We'll dive into how to scale AI products that governments
1:22
actually trust, the tightrope between privacy and security,
1:26
and how empathy and resilience shape great companies.
1:29
If you're a founder building an AI, SAS or security tech, this
1:34
one's going to light up your brain.
1:36
But for some business. This episode is brought to you
1:38
by Kitcaster, the podcast booking agency that gets
1:41
founders, execs and entrepreneurs on the world's top
1:43
podcast. Imagine spending an hour sharing
1:46
your story and walking away with thousands of people who now like
1:49
you, trust you. That's what podcast guesting
1:52
does, and Kitcaster makes it easy.
1:54
We've booked over 15,000 interviews for more than 500
1:57
clients and our founders have raised hundreds of 1,000,000,
2:00
landed massive partnerships and built unforgettable brands all
2:04
by telling their stories on podcast.
2:06
If you're ready to step up your thought leadership and turn
2:08
conversations into conversions, head over to kitcaster.com and
2:13
apply today. That's Kitcaster by MO burst.
2:16
What's up everybody, let me hit you with some math that will
2:18
blow your mind. A wealth manager using warm
2:21
start dot AI grew their assets under management by 8.5 million
2:26
in three months. Their monthly fee?
2:28
500 bucks. That's a 1700% return on
2:31
investment in three months. And I get it.
2:34
Another AI tool, another monthly subscription.
2:37
But here's what makes Warm Start different.
2:39
It's not trying to find you new people to bug.
2:42
It's helping you reconnect with people who already know you.
2:45
As a founder, your biggest asset isn't your product or your
2:48
website. It's your relationships.
2:50
But most of us are terrible at nurturing them systematically.
2:53
Warm Start connects your e-mail, identifies warm contacts, and
2:57
helps you craft personalized outreach.
2:59
That doesn't sound like a robot wrote it.
3:01
The best part? You review every message before
3:04
it goes out. No automation without
3:06
permission. If you're ready to stop leaving
3:08
money in your inbox, try warm start dot AI.
3:12
Use code AIFF for 50 bucks off your first month because the ROI
3:18
on your existing relationships is probably better than any
3:21
investment you'll make this year.
3:23
That's warm. Start dot AI and use code AIFF.
3:35
Hey, I I take the flight in the globe with me on Brian.
3:45
All right, we're rolling. Galvin, How are you?
3:48
Welcome to the show. I it's great to be here.
3:52
Yeah. Appreciate you inviting me over.
3:57
It is my pleasure. Now give us a quick introduction
4:00
and overview of your company, Loretta IO.
4:03
So, yeah, so Loretta IO is a computer vision company that
4:09
started around along this path in 2019, seeing the first
4:15
Transformers in the wild and seeing an infinitely scalable
4:19
network of computer vision. And what we wanted to do was to
4:23
build a system that could describe and understand the
4:28
physical world through the lens of humans.
4:32
And So what our system does is we identify human behaviour and
4:37
intent and we use that to track people in retail and security
4:41
settings. Amazing.
4:44
So is this to prevent shrink or what's the application?
4:47
So in the retail setting, a lot of what we do is with shopping
4:51
malls to do leasing and marketing.
4:53
So kind of closing the Omni channel loop is what they used
4:57
to call it, which essentially means I see an advertisement, I
5:00
buy a product, make sure that the people see the
5:03
advertisements, buy the product. And in security, then it's a bit
5:07
wider than that. So when security, we work with
5:11
TSA and Homeland Security and special operations company
5:15
countries in a couple of countries.
5:16
And essentially what we do is we anticipate people with malicious
5:22
behaviour so that we can look out for them and kind of get
5:26
them a kind of AT -1 identifier of when something is about to
5:31
happen. Unbelievable.
5:34
OK, so the the the the business model is is built for security
5:39
and retail as far as like reducing shrink, reducing theft,
5:42
reducing any kind of crime intention.
5:45
The the the purpose is to like stop problems before they.
5:48
Start That's exactly right. And I think the way that we
5:51
think about it is hundreds of AI solutions that can detect a
5:56
crime happening in a real time. And can you build build the
6:01
system that can detect the crime about to happen a minute or two
6:05
beforehand? Because that changes the
6:08
narrative, both for there would be criminal but also for the
6:11
security people involved. Amazing.
6:16
OK, I could see that this is high need high demand.
6:21
Talk to me about the technology and exactly what AI and what a
6:25
is role is playing with AI vision.
6:27
Yeah. So this, I think this is a very
6:30
important thing to understand and I think it's also the the
6:36
driver of how this business has lasted for so long.
6:39
So in 2017, 1819 when we kind of initially incorporated, there
6:44
were about a dozen computer vision companies in Singapore.
6:48
They were all AI4, XY and Z. And currently we are one of two
6:53
that are still alive and the other 10 have disappeared.
6:56
And well, the reasons for that is the way that we taught about
7:03
computer vision, right? So there are a lot of AI
7:07
solutions out there, which I kind of call TIN AI by TIN AII
7:14
mean the product itself is the output of the algorithm.
7:19
So the algorithm detects fight. It has fight detection AI and
7:23
great. And the thing about site
7:25
detection AI that we saw then as a practitioner, but has become
7:30
more and more true every year, is all of the work that you have
7:35
done to make your AI 5% better than open source and gather your
7:38
data set that is 5% better than the public data set.
7:42
All of that improvement was completely superseded by the
7:48
thing that is GPT or the future model or the new AI model, so
7:53
that a person coming in two years later with none of your
7:58
information and data and background could build a system
8:02
that is comparable to the system that you build.
8:04
And so instead what we did was we believe that to build a
8:09
sustainable business, the AI has to be a foundational component.
8:13
And So what is the technology behind Loretta?
8:17
It is not that we detect anything better than you, it's
8:21
that we first re identify. So that is to track a person in
8:26
one camera is the same person as is in the other camera, non
8:30
biometrically, which means I see you from the back.
8:33
If I see your pants, I can you turn around and see your shirt
8:36
and say this is the pants of the, this is the shirt of the
8:39
guy who was wearing that pants previously.
8:41
And so by doing that, we take that as step one, Step 2, We
8:45
take that and build it into what we call an entity model, which
8:48
is in a building with 300,000 people.
8:51
We have 300,000 end to end narratives of a journey.
8:56
So 300,000 people taking a journey through a building.
9:01
And so we take that journey and then we use AI to augment that
9:06
journey with more and more data on your behaviours and think
9:09
that you do and think that you shop that please how they shop
9:11
that because they looked at the people they'll be they are with.
9:15
And then we build our services on top of that data.
9:18
By doing that, we've decide we, what we know is 2 years ago we
9:23
have the best Reid or tracking system.
9:26
Maybe 2 years, maybe five years later, they'll be that system
9:30
will become widely available. But if a person comes in then
9:34
with that system, they don't have this, the information and
9:38
the knowledge built on top of that that could not have existed
9:42
if you didn't have the system earlier, right.
9:44
So that's kind of the way that we have looked at things.
9:48
Incredible and he he said when was the company founded the.
9:52
Company founded in 2017 as a full service.
9:56
That IoT business became a full of pureplay AI business in 2019.
10:03
Well, amazing now in that we're 2025 now.
10:07
When does when does the timeline for Minority Report kick in?
10:12
So as the company who has investors who think about stuff
10:20
like that, I cannot give you too much detail about it.
10:23
But one of the things that I can tell you is that we built the
10:31
company with this philosophy that even there are 5-6 years
10:35
ago, we thought like, OK, everybody's going to turn us
10:38
into the Minority Report company.
10:40
We kind of have all the capabilities to do exactly that.
10:43
The only thing we don't have is the tree blue ladies floating in
10:46
the water and the bag. We only have one lady so far.
10:49
We need to get another two. But essentially what we decided
10:54
was you can, if we our company succeeds, it will decide how
11:02
surveillance is done, or at least how much surveillance the
11:08
world is going to accept, right. And so we built a solution that
11:13
is end to end 100% non biometric, that is to say is
11:19
built in such a way that until you cross the same line that you
11:24
already crossed, if you're a criminal, and then the police
11:27
can take action against you. If you don't cross that line, we
11:31
will give the government or the police force no more data than
11:36
they already have. Another words.
11:38
Everything about you gets removed until the moment you are
11:42
a criminal. OK, maybe once you have crossed
11:44
that line, maybe your experience will be way different, right,
11:48
Right before that the for the rest of us, let's do nothing.
11:53
And I think that that was a very important decision that we made.
11:57
It was a decision that we made that made us lose battle for
12:02
contracts throughout every single year.
12:05
But I think over the long period will allow us to win the war.
12:08
That is kind of major narrative behind that.
12:13
Yeah, how long do you keep the data?
12:16
24 hours. Really.
12:18
So it's quick. Yeah.
12:21
You know, it's interesting. It reminds me a couple of
12:23
things. Minority Report I I is low
12:25
hanging fruit. Of course people don't know it's
12:27
a movie that takes place in 2054.
12:30
So it could be spot on where they're able to have what was it
12:34
called Precog, where the the cops would know you're going to
12:38
create a do a crime before you do and they'd arrest you first.
12:42
Also reminds me of a great Radiolab podcast episode where
12:46
they talked about I think it was I believe in Fallujah where they
12:50
just basically flew drones 30,000 feet above the whole city
12:54
and just took video of everything.
12:57
So then what what could happen is now essentially you have a
12:59
time machine that can go back in time.
13:02
So if you see, for example, like if there's a, you know, a
13:05
roadside bomb that goes off, well, they just start rolling
13:08
back the camera so they can see, OK, last night we see characters
13:13
that put the that planted the bomb in the road.
13:16
And then you go back further and you say, oh, OK, they're at this
13:19
house and they met with these four people.
13:21
And then you can start to then you speed up the camera to be
13:24
like, OK, well, where did they go afterwards?
13:27
And they can start tracing them down, you know, so when we're,
13:30
and, and, and this is like fascinating stuff, but it
13:33
really, really creeped people out.
13:36
Like they, they decided that they're like, I don't think we
13:38
want, you know, this kind of fidelity.
13:41
It, it could be a, an intrusion of privacy.
13:44
So when you're, you're thinking about that, how, how do you see
13:48
like the, the, I don't know, the, the temperature of public
13:53
opinion to kind of embrace some of these, you know, a little bit
13:57
more intense AI driven surveillance techniques.
14:02
It's. Such a difficult question
14:05
because the problem with these systems is that you only need to
14:12
have a few inroads and then it gets normalized.
14:16
And then essentially you can, you can change public opinion
14:21
one step at a time, kind of like how Uber made everybody
14:25
comfortable in walking into entering people's cars over the
14:28
last 20 years, right? And so there are always
14:33
locations that are willing to do this.
14:38
Now, living in Texas, where I Live Today, I live about half an
14:43
hour, half an hour away from the only, I'm not going to tell you
14:47
the name of it because I don't want people to go with this, but
14:49
the only school district which allows teachers to open carry.
15:00
OK, Texas is wild. And I also, I also know one of
15:10
the security people who have like, we also admit that at
15:12
least nine of those ten teachers are going to panic if there's
15:15
actually going to be a need to use that gun.
15:17
And they're going to like, probably like shoot themselves
15:19
to the foot. Like there's no way that this is
15:21
a great idea. But there are always pockets of
15:25
sentiment that will normalize these things.
15:28
And then you have to make a decision how to move forward
15:30
with it. And the way I think the the way
15:34
that we've seen it is that for better or for worse, it is
15:38
getting normalized. And the part that I'm far more
15:42
frightened about, it's less about the surveillance because I
15:46
see surveillance or the information coming into the
15:50
network of these things of like knowing behaviours or things
15:54
that are happening in the space as often being a net benefit.
15:59
And the reason for that is whenever a police officer stops
16:03
a vehicle, and if you are driving this, this has happened
16:06
to all of us, right? Police, the officer has stopped
16:08
the vehicle. And if you take 100% of what
16:12
they have in their mind is stopping you, I would probably
16:15
say it's like 20% what they saw, 5060% their preconceived
16:21
notions, which are where things like stereotypes and biases.
16:25
And this is the place that I am working on today and 20% how
16:30
they feel today that this is kind of what's making up this
16:33
100% desktop tube, right. And as the amount that you
16:39
provide to them increases that 10% of the edge, which is the
16:43
objective truth, expands and removes the space for your
16:48
stereotypes and these things to come in, right.
16:52
It is more like the framework in the mind of the investigator who
16:56
is deciding whether you're innocent or guilty versus the
16:58
person who stopped in the street.
16:59
That that's kind of the way I think about it.
17:01
The thing that is far more frightening to me is where AI is
17:04
making decisions. And that is the part where yes,
17:10
maybe there is pre caught here data that might be available in
17:15
certain parts of the world, probably not in the US, probably
17:19
not today, but in certain parts of the world, yes.
17:22
All the things that you saw with the drones flying over, this is
17:26
something that we have been doing all over the world outside
17:29
the US for ages, right? The satellites in the sky and
17:34
you saw images of like, OK, if you think of like in our era, if
17:38
you look at like the WMD narrative in the Iraqi before
17:42
the prior Iraqi war, you they showed the satellite images that
17:45
said like this is what I saw this tank, this tanker is
17:49
probably driving a bomb, etcetera, etcetera.
17:51
Well, we do know now that it's not true, but there was the data
17:54
was being captured, right. And so we know that that's
17:57
there. The question is, how are we
17:59
making decisions about it? And the question is, do we
18:04
become good at making decisions about it faster than we give the
18:08
decision making to the algorithms?
18:10
And I am not that comfortable right now giving decisions to
18:15
the algorithms. Like, I think that there are
18:18
ways to squeeze the hallucinations, but there are no
18:21
ways to eliminate them. And sometimes people's lives are
18:25
at stake with these decisions, right?
18:26
And I'm not quite sure whether we are ready for that yet.
18:30
And that might be the concern that folks have, particularly
18:33
with AI it because we've all accepted if you're going to the
18:35
mall, you going to the airport, you know, you're on camera, you
18:38
already know that, you know, we're going into an age now
18:41
where all of our glasses will be wearables that will have will be
18:44
recording everything. Everything will be recorded
18:47
everywhere, all times. We're walking into that kind of
18:50
reluctantly, but the benefits are, are so great that it, it's
18:54
just a matter of adoption, you know, But perhaps what people
18:57
are upset about is if you're talking about bias, if you're
19:00
talking about bias with the police force, you're talking
19:02
about bias with security, you're talking about stereotypes,
19:05
you're talking about, you know, basically you're talking about
19:09
like prejudice against a minority, you know, and the, the
19:14
problem, I guess is, is that AI from, you know, its inception.
19:18
I'm not even say AI. I'm going to say once GPT really
19:21
started getting rolling in 2022, then people are like, dude,
19:25
these, this machine is insanely biased.
19:27
You know, like in the beginning, if for example, they were, they
19:31
were, if you put in a ask for a picture of the founding fathers,
19:34
they'd be like black guys, you know, so, so people are probably
19:39
a little bit worried about what's the bias innately built
19:42
into the AI? What's the training they're
19:44
receiving? You know, so maybe you can speak
19:47
to that a little bit. Like how how do you fish out the
19:50
bias in the machine? Now, you and I know startups are
19:53
built in board rooms. They're built in the wild during
19:56
3:00 AM, bug fixes, botched launches and that one meeting
20:00
where everything almost fall apart.
20:02
That's what Code Story is all about.
20:03
It's a podcast where tech founders breakdown the make or
20:06
break moments behind their products.
20:08
What went wrong, what went right and what nearly wrecked
20:11
everything. Hosted by Noah Labhart, Code
20:14
Story goes past. The highlight reel and into the
20:17
honest stuff, late pivots, brutal bugs, and somehow coming
20:20
out alive with a product people actually love.
20:23
If you're in the build stage and the scale stage or the please
20:26
Let this feature Work stage, I can relate.
20:29
This show will speak to your soul.
20:30
Listen to codestory@codestory.co or wherever you get your
20:34
podcast. Just search Code Story.
20:36
It'll be right at the top. Because great startups aren't
20:38
born. They're forged one hard moment
20:41
at a time. Code story.co.
20:46
This episode is brought to you by Moburst, the AI first growth
20:49
engine trusted by iconic startups like Calm, Robin Hood,
20:53
Reddit, Uber and Google. Moburst is a digital full
20:57
service, mobile first marketing agency that helps companies to
21:00
scale and become category leaders.
21:03
Whether you're finding product market fit or ready to
21:06
accelerate growth, Moburst delivers creative, data-driven
21:10
solutions that help with scale and efficiency and impact.
21:14
At Moburst, we're not just guessing what works.
21:16
We're predicting it. As an AI first agency, Moburst
21:20
uses cutting edge tools to analyze user behavior, adapt
21:24
creative in real time, and cover hidden growth opportunities.
21:28
The result? Marketing that's faster,
21:30
smarter, and built to convert from early stage startups to
21:33
global brands. We've helped hundreds of
21:35
companies overcome complex challenges, unlock new
21:39
audiences, and optimize every metric that matters.
21:42
At Moberst, we believe that creative thinking, advanced
21:45
technology, and data are the core drivers of success at
21:49
scale. If you're serious about scaling
21:51
your business with precision and purpose, it's time to talk to
21:54
Moberst. Visit moberst.com and let's
21:57
build something category defining together.
22:01
So I, I am of the belief, and this is something that is not
22:06
consistent throughout the kind of AI predictionable my, I am of
22:11
the belief that you cannot remove buyers from the machine.
22:17
So one of the reasons why I like to come on these public
22:22
platforms to talk about this noun is that today, if you, for
22:28
example, type Darth Vader into like check GPT, right?
22:32
Like, sorry Disney for using your example, but if you you put
22:37
Darth Vader's GPT, what is happening is that you will first
22:41
generate a Darth Vader like like outcome.
22:44
And then it will that there was this like thin layer of
22:48
censorship over the top that says, oh, this is a copyright
22:51
property and it gets removed, right?
22:53
And so if you say something like long clay, dark cloak, mask
22:59
wearing, space magician with a lightning sword, suddenly you
23:04
get something a lot like Darth Vader, right?
23:07
So you have bypassed the censorship.
23:09
But clearly that image was in that mind, right?
23:13
And so now what you're doing is you have to build it and you
23:16
don't know how to get rid of it. Now you have to build a sensor
23:18
layer over the top, right? That's kind of what's happening.
23:21
I think that at least in the physical world, that data
23:25
doesn't that whole data block, the pile of people in the AI
23:29
space will know, the pile of common crowd.
23:33
These are the large public data sets that are the base of a lot
23:36
of the GPT models. They don't exist yet.
23:40
And to me, you have to get rid of all the ethnicity data, all
23:44
the biometric data before it goes in, or else you will always
23:50
be finding ways to get rid of it.
23:52
One of the important things to remember is that even if the
23:55
system is designed to be objective to today, what you are
23:59
also doing is you're freezing the state of stereotypes and
24:04
bias to today, the day the data set was first generated, right?
24:09
So if let's say we have one people group hypothetically that
24:16
came into the US maybe at a later stage or still finding
24:19
their way, right? Maybe in 10 years time outside
24:25
AI, they have found their way, they are now 100% equivalent and
24:29
so on and so forth. But the AI was trained today.
24:34
And so the bias is designed in this moment in time and this
24:39
moment in time. And we know looking back at
24:42
history that every time we look 20 years into the past, we're
24:45
like, what were those people thinking, right?
24:47
And so we don't want, we don't want the entire future to be
24:51
looking back at 2025 and thinking about us as like the
24:55
morals who not only were like this, but also froze our
24:59
perceptions into our systems, right?
25:02
And so this is kind of, this is the only way I think, I don't
25:05
think you can ever remove the system because it's pattern
25:09
magic, right? And so stereotypes are also a
25:11
pattern matching system. And so the same logic that makes
25:14
it complete sentences, it's the same logic that makes it think
25:17
that way about the people as it does today.
25:21
Yeah, there's something unsettling about that, but it's
25:23
so common. You know, it's obviously it's
25:26
horrible when it, if it comes out in like racist ways, but
25:29
like it, it comes out in, in productive ways all the time.
25:32
Particularly I'm thinking like if I was a head of security for
25:36
like a major retailer. You know, I, I just looked it up
25:39
real quick and 30% of shrink from retail comes from employee
25:44
theft, you know, so that's just data.
25:47
So that's not saying that's a bias against your own staff when
25:51
you know, a vast majority. And that 30% is really half,
25:54
because it also says that the other 30%, another 30% is
25:58
basically administrative mistakes.
26:01
You know, so you have basically of all the people stealing stuff
26:04
in your store, half of them are employees.
26:07
So, and we're also accustomed to this.
26:09
You see oftentimes that the cameras are actually on the cash
26:12
registers. It's not following people
26:14
around, you know, now there's no date on this, but anecdotally, I
26:19
would say a good percentage of outside theft is coming from
26:22
teenage girls. Like, I mean, I remember in high
26:25
school the the all the sweet, innocent teenage girls I met
26:30
were like absolute kleptomaniacs, you know, so and
26:34
so I would have liked if I'm. Had a gift?
26:39
At some point, I'm head of security and a group of teenage
26:42
girls comes in. I want the, I want all the like
26:45
alarms going off and going crazy to make sure we keep an eye on
26:48
them. So there's certain inherent bias
26:51
that you're, we refuse to like allow for good reason.
26:55
But then there's other bias that like we probably encourage.
27:00
So if, if your job is to say that the machine will never get
27:03
the bias right, we're only at as good a point as we are right
27:07
now, how do you kind of walk that line and how do you reset
27:11
your data points to make sure that the biases we accept and
27:15
don't accept are updated? Yeah.
27:16
So I, so, so this is such a good point.
27:22
One of the reasons why we think we're a pretty good company to
27:26
be doing this is that we have a background that's very different
27:29
from a lot of AI companies. And so me and my brother, before
27:33
we ran, before I ran Loretta, I also ran physical restaurant
27:38
chains and we used to run restaurant chains that work
27:41
specifically with the migrant population.
27:43
I'm going to share a little a secret anecdote that is well
27:47
known only in the in the, in the in the restaurant industry,
27:51
which is that often when you want to look for the very best
27:54
chefs, you go to parole centers. Yes, parole centers.
28:01
Parole centers. So like people like just got to
28:04
go. Yeah, That is where you get the
28:08
best cooks because everything that you saw in Kitchen
28:11
Confidential is true, but it has a dark side.
28:14
And so when they when they hang out and take drugs together, it
28:18
also means that if one of your old employees want to complain,
28:21
everybody goes to jail. Yeah.
28:27
Amazing. Yeah, man.
28:29
Hey, I know a lot of chefs and I don't know if they're convicted
28:32
felons, but they are felons for sure.
28:34
Definitely felons. And so one of the things that
28:38
happened, so we bought them. I used to go in the migrant
28:41
community. This was in Singapore by the
28:42
way. And Singapore is a place where I
28:45
think Singapore's drug laws are legendary in the world, right?
28:48
So essentially about this much is cocaine.
28:53
It's enough for the death sentence in Singapore.
28:54
It's about 9 grams is enough, right?
28:56
So it's pretty bad, but we had somebody who he still had his
29:04
ankle monitor, right? So he was obviously still at a
29:06
halfway house. He was still on parole.
29:08
He was a mentor to a lots of people in our restaurant chain.
29:12
So he was a mentor to all the young staff, all the young cooks
29:15
and all the young chefs. We made it so that every from
29:20
2:00 to 5:00 PM when the restaurant was quiet, his family
29:24
was allowed to come into the restaurant to live, to hang
29:27
around there and to do the whole world and stuff like that so
29:29
that he could see his kids, right?
29:31
And so with parole, you, you actually are only allowed to go
29:34
to your office or your workplace and go back to the place, right?
29:38
Because they're tracking you all the time.
29:40
You cannot leave that particular pathway and stuff for that.
29:43
And so we did a lot of these things.
29:45
And the reason why I mentioned this is that we were, I want to
29:51
separate a, this is an ex convict.
29:55
So if there's a person who is a person of interest by default or
29:58
by stereotypes, he is it, right? He is.
30:02
He had 10 years in prison. But when he's out, we want to
30:06
build a system that does not look at him more carefully than
30:13
it does other people. So it's not about.
30:15
So once that happens, you kind of fall back into that same
30:19
problem that we have, which is like, oh, we talk about how like
30:23
in a black community, a white criminal often never gets caught
30:27
because like all of the police people are just like looking for
30:31
the, the black community that the people there and they were
30:34
trained that way. And it's like, and that skews
30:38
the ratios even more, right? It's like, maybe that is,
30:42
there's some data that implies this, but we don't want the data
30:47
to drive the decision. And so the way to do that is you
30:50
have to make objective capabilities that are narrow
30:55
enough that the things that we want to avoid.
30:59
So racial bias, age bias doesn't get involved in it at all.
31:04
So yes, maybe I will train the system that it knows how to do
31:08
test for a person in a coat or putting stuff in a handbag like
31:14
yes, handbags are tune more towards younger women versus men
31:19
and so on and so forth. And so obviously that
31:22
capabilities in there. But if I I don't want the system
31:25
to have what is what is called in the AI space today when
31:30
called set detection, a set SET means a detection that says when
31:35
a person like this does that, then both of these things are
31:38
detected together. And that's where it gets
31:41
problematic. But then you are codifying the
31:44
stereotype into the system. Yeah, it's tough too, because
31:48
sometimes statistically isn't helping those stereotypes.
31:51
You know, it's kind of it's tough.
31:53
It's it's, it's, it's not an easy not to untangle honestly.
31:58
And it probably takes a lot of bravery to to even give it a
32:02
shot, you know, I mean it. And we kind of go back and forth
32:06
to your point, you know, after COVID, like the take grocery
32:10
stores for example, a lot of them basically just bailed on
32:13
the self checkouts because they were getting robbed blind.
32:17
You know, like people would just come in, throw some M&M's on the
32:20
thing, pay for that and throw 3 steaks into their bag.
32:23
You know, or they're doing the same thing at like the the Whole
32:27
Foods over by Wash Park where I live.
32:29
You know, they basically said that none of the high school
32:32
kids could even come in there any.
32:33
That's right, there was. A sign that said that if you
32:37
were under this age you cannot use self check out.
32:39
Say what? You can, yeah, you cannot shop
32:42
here, which is which is wild, but I mean, this is what
32:44
happens. The line moves, you know, so the
32:47
things that we're willing to accept will change depending on,
32:50
you know, the outcome of the scenario.
32:53
It's very interesting. So you mentioned you ran
32:57
restaurants. How do you, Tran, how do you,
33:00
how does your career switch from running restaurants in Singapore
33:04
to, you know, carrying a gun in Texas?
33:07
So it was a short jump, right? No, I'm kidding.
33:12
But essentially the way they used to work is I had a day job
33:18
that when there has AI have a very strong data analysis
33:24
narrative that goes through my life.
33:26
And so it went, management and consulting and Monte Carlo
33:30
simulations, all that kind of stuff.
33:31
And they went into banking and big data.
33:33
And while I was doing this, I ran restaurants for a full 6
33:37
year period at the same time. And I was doing all of that.
33:40
And the way that I looked at it is that I decided to be.
33:46
I think everyone who becomes a startup entrepreneur looks at
33:51
things one of two ways. They either think of himself.
33:54
This is a great, there's a great gap that I can fill or there is
34:00
a fundamental large gap that needs to be filled.
34:03
But is this a quick win or is this a slow burn kind of
34:07
solution? And I was dumb enough to go for
34:11
the slow burn. And so essentially the way I did
34:16
it was that when I looked into this, I looked at the way that
34:22
physical data was being captured and I didn't see humanity in the
34:28
data. And what I mean by that is not
34:32
like is that if I, we were looking at AI or computer vision
34:37
of in physical spaces in 2017, 1819, even today, most of the
34:41
time you'll see here the words like footfall, dwell foot, words
34:46
like GMV, These are averages of people in space.
34:51
It's actually more describing the location than is the person.
34:56
And then I, I took that narrative.
34:59
I went to a couple of the major shopping mall providers in
35:02
Singapore, vendors, shopping mall orders in Singapore.
35:07
And I was like, there has to be a secret to how you are
35:09
unraveling this data. And I learned that a secret is
35:13
that they're not unraveling the data.
35:14
So they're just guessing. It is the reason why a lot of
35:19
the top mall operators in the world at the same mall operators
35:24
choppy from mall to mall because it is all heuristics and advice
35:28
and not data that is driving rental and leasing and shape and
35:33
design decisions of these malls. And I was like, that's dumb that
35:39
and that's problematic. And so I built a technology
35:44
specifically to solve the solution of bringing humanity
35:48
back to it. Because every footfall data
35:52
doesn't tell you why they are there.
35:54
Are they? Just because people walk by a
35:56
shop doesn't mean that the shop is had a lot of business.
36:01
It could be a lush. It smells amazing.
36:04
Like. First, like, OK man, like that
36:07
doesn't mean anything. We need data that we can use and
36:11
that is all about the what is in the customer's brain and not in
36:15
their movements. And so I was like, what
36:18
information could allow us to identify the identify the intent
36:22
of the customer? Funnily enough, what I ended up,
36:26
what we ended up building, we found that identifying malicious
36:30
intent is easier than define shopping intent because shopping
36:35
intent is very nuanced. It's like I bought these four
36:38
things, but I wanted to buy these other things.
36:40
I saw this, I could have paid more, but I didn't.
36:43
And a lot of this is subconscious to you.
36:46
Malicious intent is almost entirely conscious, and so the
36:50
security product is an offshoot of the retail problem.
36:56
Basic and it might be just the the entry point for all this,
36:59
because as the fidelity of the data gets greater and greater
37:03
and we have more data points, then there will be more buying
37:06
intent as a that that's kind of a secondary problem.
37:09
So you have this great idea, you go to the the mall operators,
37:14
you realize that they have no idea there is an entry here.
37:17
So did you raise capital? How did you finance the the
37:21
company and it's in the beginning.
37:22
Oh, we, I went to the mall operators.
37:26
I first got I have a bit of a consulting background, so I
37:30
first got a contract to analyze their data.
37:34
So this was my foot my this was how I got into the door
37:38
understanding how the data worked.
37:41
That is my version of validating product market fit, right?
37:45
Because there are different ways that you validate, especially if
37:48
you're in the B to B space, you have to know that there are many
37:53
gaps, but you don't from the outside, you may not know
37:57
whether the gap is real. So you might see, let's say if
38:01
you go into a car into like a car dealership, it's like, OK,
38:04
they seem to not be taking me seriously.
38:07
I'm sitting in the corner, nobody dealing with me.
38:09
There might be a gap. It's like, no, there is a reason
38:12
why they're ignoring you. They're making you sit.
38:15
There's a reason why I trust that they don't want you to get
38:18
out faster, right? So they're kind of things they
38:20
need to understand and I need to validate these things.
38:23
And so I built that first. Once I got my first contract
38:26
there, I told that money I got into a property type accelerator
38:33
where I give away 2% of the company for no better reason
38:37
than to have a pilot at a property mall in Singapore.
38:41
So at a mall called Lend Lease, there's a mall operator called
38:47
Lend Lease. Lende's, by the way, is pretty
38:49
famous in the US. Most people don't know the name,
38:52
but it's the company that had the contract for the
38:54
revitalization of San Francisco over the last eight years.
38:57
So. Wow.
38:59
The contract that they did not fulfill, but there was a
39:02
contract that they had. Pretty big job if we're talking
39:07
about San. Francisco.
39:08
We're talking about San Francisco.
39:09
It was a big job, but they that we got them as a investor, they
39:15
put us into their mall. We validated that and then once
39:18
we had the validation and we knew that a product had value,
39:22
then we raised money. We raised money in Singapore,
39:25
which means we didn't raise a lot. 1 of the problems with
39:29
raising money overseas and coming to the US is that the
39:33
quantums are very different. And one of the main reasons for
39:36
that is that every time you raise money, you have to have a
39:40
decision of where the money takes you, right?
39:42
So it's like, do I go from research to pilots, from pilots
39:46
to scale up? By the time I raise the next
39:49
round, am I going to be showing milestones or am I gonna be
39:53
showing graphs? Right?
39:55
So it's kind of two ways to do it.
39:57
If you talk about a seed, you can go from AC to a Series A on
40:01
milestones, which is like, OK, I got my FDA approval, I got my
40:05
security approval. I got a pilot in a government
40:08
location. I got a pilot in the New York
40:12
Trade Station or something like that, which is great.
40:15
It doesn't have enough that that doesn't show that you have
40:17
enough revenue, but it shows you're going very quickly in the
40:20
right direction. So it's speed against
40:22
milestones. Then that's the next group.
40:25
That's the next stage, which is when you go to graphs and graphs
40:27
is where you just have to show revenue did this and sales to
40:30
test and you if you can, if you can, if you once you get to
40:34
graphs, then it's all performance, which is great for
40:37
companies that are building deeper and longer term.
40:42
Sometimes you have to take, you have to step away from the take
40:48
up graphs to move yourself deeper into the temp.
40:52
So that I have a product that's a $10 million temp and I'm going
40:56
to scale it up into a product of $100 million temp, then I'm
40:59
going to scale up to a product of a billion dollar temp.
41:02
And so because we needed, so we decided that we wanted to go
41:06
deep. And so we wanted to make small
41:08
jumps deeper into a scaled and more mature product and do that
41:13
while we were going down at the same time with an AI product.
41:17
One of the key things that you need to realize that you're
41:20
moving in a space with two problems that a lot of the SAS
41:24
world, they don't need to worry about 1 is that there are a lot
41:27
of other AI providers out there. And so your competitors are
41:32
fast, right? A lot of all, a lot of
41:36
successful start-ups. There's a famous saying by I
41:40
think Paul Graham there that says, Paul Graham used to say
41:44
it's great to work in industry where your competitors are slow,
41:48
right? And so in this, when you're in,
41:50
where you're in AI, your competitors are fast.
41:53
So that's one problem. The second is that your
41:55
technology, the technology is evolving fast and it's evolving
42:00
in a way that the way I like to talk about it is it thickens, is
42:06
coverage of your service. And So what I mean by that is an
42:11
AI that only detects guns, for example, is not a full suite of
42:17
a security solution. But if your AI today can
42:20
describe a location and run an agent, an agentic model that
42:25
calls the police or something like that, then the entire
42:28
solution is in the AI wrapper. If your entire solution is in
42:33
the AI wrapper and all of that is provided by a third party
42:36
solution like Open AI or whatever where you're just
42:39
sending in the images, then your own differentiation is shrinking
42:43
fast against a new incumbent. If you do that, then your
42:48
solution cannot last. And so if you're going to go go
42:51
and make quick a quick bite out of it or you're going to sell
42:54
your service as a component to a big provider, then yay.
42:57
But in all other cases, you are losing relevance every day as
43:02
the solution becomes, as the core wrapper expands its
43:06
capability. We never had this in the past
43:09
because in the past every API never left, kind of never got
43:15
out of its lane, right? So if you use Twilio, it would
43:17
be giving you SMSS, but it won't be putting in the content of the
43:20
SMS. So you still have your system
43:22
that sends you the messages and so on and so forth.
43:24
AI doesn't do that. And so we built a solution
43:27
that's like that. We we built in such a way that
43:29
we could build differentiation that expands over time, even if
43:35
the AI gets wider over time. And that those are, I think,
43:40
fundamental in the AI space today.
43:42
So yeah. It would.
43:44
Did that influence the way you're able to raise or was that
43:47
kind of seed raised in Singapore?
43:49
Was that the the last check? How did that, how did the
43:52
financing kind of progress? From once we had that check, we
43:57
were able to do 2 things. One is that we were able to
44:00
differentiate ourselves in the industry.
44:03
The next check after that was from a strategic investor in the
44:07
US. So we continue to do that and we
44:10
are going to continue raising in the US.
44:14
The second thing that we that allowed us to do is that allowed
44:17
us to read ourselves and kind of scale up there.
44:20
And I think one of the important things of AI, or at least with
44:23
any emerging technology is we've, I believe that if your
44:27
technology has is going to scale, you need to always have
44:31
one hand in the cutting edge and in AI or in the robotic space
44:37
that cutting edge is right now in government, in defence, in
44:41
military use cases and so on and so forth.
44:44
Having one hand there forces your technology to develop.
44:48
A lot of the reason why our competitors in this property
44:52
analytic space stall is because once you reach there's a point
44:58
where your product has to match the requirements of your
45:02
customer obviously. And in many cases, the customer
45:06
does not want your AI to improve.
45:08
They just want your system to be more enterprise and cheaper.
45:10
And so if that is your case, then the problem is that because
45:16
AI is still improving, you're going to build for that
45:19
particular customer and then a new startup will be able to do
45:24
your capability better than you because they were still
45:27
investing in the algorithm in the past.
45:30
And with a, with smaller razors as well as the razors that you
45:32
do overseas, you can't do both at the same time.
45:34
We don't have to budget for it. And so there's a bit of a
45:37
balance that you need to make between those things.
45:41
Looking into the future, what would be a good data point for
45:45
the software? Right now you have video, you
45:47
have audio. Like what if you could be able
45:50
to see people's body temperature or how much they're sweating or
45:53
like their eyes looking shifty? Like what would be a good
45:57
additional biometric? So we've looked into this more
46:04
than I like, but it's actually the answer is that there isn't a
46:13
lot more that we need to do in the computer vision sets.
46:20
So the second the next area is will likely be around stand of
46:27
EMF sensors. So what I mean by that is
46:29
essentially terrafant sensors and millimetre wave sensors and
46:33
so on and so forth. They are low cost.
46:35
They are starting to go longer distances and they're starting
46:37
to get very good data coming in. And that is essentially the
46:43
X-ray vision version of it where it's like, OK, I see you
46:47
removing object from the shelf and I now can use the the stand
46:52
off to see you putting it in your bag.
46:54
So that's kind of the other system they're coming in and
46:58
there is a contract that we have that's publicly known.
47:01
And there are a couple of these that are going out up there
47:03
where the computer vision that we use becomes the agent layer
47:08
for processing of other data. And by agent layer, I don't know
47:11
what agent has kind of been taken over by the AI world to be
47:15
in APIs. But agent in the old school days
47:20
of machine learning means that you attribute all data to the
47:24
individual. And so it's like an individual
47:27
that walks by an X-ray, which that walks by a millimeter wave
47:30
sensor or a Geiger counter even, right, which is for radiation.
47:35
Our system will then say this, when this person entered the
47:38
this area, the Geiger counter went take that.
47:42
And when you left it went down. And that means this guy, the guy
47:46
who has the nuke in the bag. Yeah, he's got The Dirty bomb.
47:52
Tackle him. Release the dogs.
47:53
Release the dogs. Wow, so fascinating.
48:00
I really appreciate your time. We're going to do the lightning
48:02
round. Are you ready for the lightning?
48:03
Round. I love that you kind of put the
48:09
company on your shoulder. What what does founder LED
48:12
marketing mean to you? Why?
48:14
Why do you come out here and and talk on podcasts about your
48:17
company? Founder LED marketing to me
48:19
means that you embody the vision of the product.
48:24
They embody the way that people will perceive the company in
48:28
terms of how they will trust you and how they perceive you.
48:31
And so having like an ethical narrative, having a trust
48:34
narrative, having a technical narrative means that when a
48:37
customer talks to us, they know that they're not going to get
48:40
too much marketing. They're going to get something
48:42
that we know it's going to stay and has a meaning behind them.
48:47
Amazing. I love it.
48:48
Who would be your dream client? Who would you like Walmart or
48:52
something? Like what would be your a game
48:54
changer client for you guys? I think the one of the clients
48:58
that I really want to get in soon is the Simon's Mall brand.
49:04
And the reason for that is because malls in the US, they
49:09
kind of, they kind of badge so different.
49:14
Like even in the North, even in North America, you have the top
49:18
ten malls in North America, like 9 of them are going to be in
49:20
Canada, right? Like it's crazy.
49:23
And so it's like this is something that can meaningfully
49:26
transform the way that we experience the physical world
49:29
and experience social life. And so like, I would love to
49:32
have silence as a customer. Yeah, I agree. 8 malls.
49:37
They they need an overhaul. Let's say the audience of this
49:45
podcast loves ideas, business ideas.
49:49
If you What's a business that you're interested in making that
49:52
could make that would earn $1,000,000 as soon as possible?
49:57
So I looked at three of them. I look at a couple of these
50:03
every couple of weeks because I want to test out the vibe coding
50:07
stuff and learn how to code on my own.
50:09
And I, I test these out and I'll begin.
50:13
The first one is I believe that we completely that OK, So one of
50:19
the important things that one of the important data points 1/3 of
50:24
all open AI traffic disappeared in March of, in May of this
50:28
year. Do you, do you, do you hear
50:30
this? And it came back, it came back
50:32
in August. Do you know what that was?
50:34
That's called summer break. What?
50:38
Summer break? Summer break, Yeah, the kids
50:41
aren't using. They're, yeah, they're earning
50:44
their money from coders, but the people who are driving traffic
50:47
to chatgpt.com are students cheating on their homework,
50:51
right. But like that, that is the key
50:54
driver of AI today from a traffic perspective.
50:57
And there is an end to end app that can be built entirely on
51:03
your lecture that obviously is gonna work, right?
51:08
You can transcribe the lecture, you can translate the notes, you
51:13
can reference that to your textbooks and your homework,
51:18
close it out and still have a system that you can ask
51:21
questions to. Essentially, you can get your
51:23
tutorial entirely on an agentic AI model that will run
51:28
$1,000,000 in a weekend for sure, right?
51:30
So that's one of the things that clearly would work.
51:35
The second one that I think is really interesting is I like the
51:40
idea of using computer vision as a assistance tool in the
51:48
physical world. There was a interesting image
51:52
that we thought that we I saw online that I absolutely love it
51:55
was it was the parent of one of those like AI scientists and she
52:02
used chat GP to identify a book on the bookshelf.
52:08
And literally you can take that system and that means to me, you
52:14
could take your little recording.
52:16
What's your McCullough they have there and say I'm looking for
52:19
something. And if you scan the room, it'll
52:23
take the photos and it will come back to you and say, I saw it,
52:25
you missed it, but I saw it. And, and lots of things for my
52:30
parents that work extremely well AI that I've been to go into
52:35
production. But a lot of these men, they're
52:37
a lot of them. Yeah, I, I, I'm really
52:40
interested in that, you know, creating an alternate memory
52:43
timeline because our memory is so flawed.
52:45
But like, if you have a camera recording everything you're
52:48
seeing, it's going to pick up everything.
52:50
You never wonder where your keys are.
52:52
You'll be like, oh, it's on the counter where you left them.
52:55
Like, oh, I thought I looked there three times.
52:57
Oh, you did. You did, you just didn't see it,
53:00
but we saw it. Thank you.
53:04
Does that does that mean that you have kind of do you have any
53:07
interest in hardware going into? Hardware.
53:10
I have an interest in going to hardware I and I think for for a
53:16
person who is coming into the startup space, hardware is
53:20
horrendously difficult to do, especially if you are in the US
53:26
My own calculations, but I have to look into this is that if you
53:29
are AUS startup going into hardware and you cannot raise 3
53:33
to $4 million out of the bat, you are either outsourcing or
53:38
you will not make it to production.
53:41
If you were in Southeast Asia, it is about four to $500,000.
53:48
So it's about a magnitude, it's a full magnitude cheaper to be
53:52
able to do it. And so just a couple of things
53:54
to think about around there. But yeah, we are interested in
53:57
hardware. There is a lot that's a huge gap
54:01
by the way. And this is, this is something
54:03
from where we work a lot with the government and suffer there.
54:06
But there's a huge gap between capabilities of vision and
54:10
sensing that are available in research and the stuff that we
54:13
have widely available. And that gap is not price.
54:17
So it cost the same. And so it's crazy that there's
54:22
no one going into these things. Assuming that governments are
54:27
deep into this kind of stuff, do you have any idea what they're
54:30
doing, what their hardware capabilities are?
54:35
Yeah, definitely things that they have, but one of the ones
54:43
that I like to think about is I worked for a while with a
54:48
present. This is a public view.
54:49
You can Google their website and stuff in there.
54:51
But I work for a person who essentially there are only a
54:56
handful, maybe a dozen people out there who run companies that
55:00
run sensors that don't run cameras that go don't go through
55:04
the CMOS sensor. Give us the CMOS sensor.
55:08
You know, when you have a digital camera, the image goes
55:11
in, it hits a bunch of pixels that get excited by the energy.
55:17
OK, sure. Yeah.
55:20
And that that whatever gets excited by the energy that gets
55:23
translated into your pixels to get into your.
55:26
Now if you do that then the amount light that comes in is X,
55:32
but the amount of it that is making stuff jump is like one
55:36
tenths of X. If you just eliminated that
55:41
system, you would have a camera that captures about 10 times
55:47
more light. Does that does that mean it the
55:53
it's that much bigger? It's 10X as big of an image, so
55:57
you can like go deeper. Into it.
55:58
It means that you could get a clear image in the middle of
56:01
the. Night.
56:03
Wow, really? Oh, so there are things
56:08
happening out there. Amazing.
56:15
Let's see, I'll ask you one more question.
56:18
Is there anything you know now that you wish you knew earlier?
56:22
I one of the things, a lot of what I've, I think the main
56:30
thing is about fearfulness. I think fear is a core driver.
56:40
If we sit back and we look at some of the like start up
56:43
founders in like San Francisco and stuff like that, we're like
56:46
these guys are reckless AF. And there are reasons why they
56:53
can, which a lot of us though they are born in Asia for
56:57
example, who have no base, who have no company to go back to.
57:01
Like you can't do that. I understand, OK.
57:04
But there is, because the I factor is the most important
57:13
factor in the IRR, in the growth of company is how fast you do
57:18
things. In some cases, for anyone who
57:23
has been an entrepreneur, an entrepreneur in the non startup
57:27
world, your settings for risks are completely wrong.
57:33
You're far too conservative. And I am.
57:36
I was like that. And there's something that I
57:38
have had to actively build and burn out of my system.
57:42
That's something that yeah, we just have to do.
57:44
We just have to do this. Do you think the restaurant
57:48
business is conservative? The.
57:49
Restaurant business is super conservative, Yeah.
57:54
One of the things about restaurant business is that your
57:57
profit margin can be something along the lines of 0 + -, 5%.
58:04
Yeah, I well, because I was in the quick service restaurant
58:07
business, I was in the top. And one of the people who was in
58:10
the top was Subway. And Subway at a time was
58:12
dominating the world. Like five years ago, Subway was
58:15
everywhere, the fastest growing fast food chain in the entire
58:18
world. And what he said was that every
58:21
time one of the stores goes into the red because the environment
58:25
no longer supports the Subway, it takes 5 stores to make that
58:30
one stop rate even. I was like, what does that mean?
58:34
Does that mean that if you had less than five stores, you go
58:36
bankrupt? And he's like, kind of yes.
58:39
Would you like a franchise? I was like, no.
58:43
Yeah, wow. How, how do you, how do you
58:48
practice taking risks? Are you like out there bungee
58:51
jumping and stuff like that, you know?
58:53
You know, one of the things that I did this year specifically for
58:56
that is that I came from a very conservative culture.
59:00
And so I have like 100 like unrealized subconscious, like
59:06
fears in my body, like hundreds of phobias.
59:08
And so my I had been going rock climbing with my wife because
59:13
I'm absolutely afraid of heist. Me too, me too.
59:18
I could never do that rock climbing.
59:20
I was terrified. They were laughing and I was, my
59:23
hands were physically shaking and it was amazing.
59:27
It's like this has been great for my business.
59:29
Rehearse. That's awesome.
59:34
You know, I, I was looking at this Reddit video, so it's like
59:39
an inch and 1/2 on my phone and it was like this guy that was
59:44
just like stacking trash cans from like 15 feet up.
59:47
And I can still feel it now like how scared I am of heights for
59:50
him. I'm just like, well, that's
59:53
amazing. Good for you Galvin.
59:55
This has been amazing conversation.
59:57
I really appreciate it. Where can people connect with
59:59
you online? So you can find me at on X at
1:00:03
Galvin West or at my website at Loretta dot IO.
1:00:09
Tremendous. Thanks so much for coming on the
1:00:11
show. Really appreciate it.
1:00:14
Hey founders, let me ask you this.
1:00:16
Are you building with AI or getting left behind by it?
1:00:20
If you're serious about scaling smarter, faster and bolder, you
1:00:24
need to check out AI for founders.co.
1:00:28
It's your go to resource for cutting through the noise and
1:00:31
actually using AI to grow your startup.
1:00:34
Each week, AI for Founders drops a sharp, no fluff newsletter,
1:00:39
dives deep with expert guests on the podcast, and hosts monthly
1:00:43
workshops where you can get your hands dirty with the real tools
1:00:46
and strategies. Whether you're precede or Series
1:00:50
B and beyond, launching something new or just staying
1:00:53
sharp, AI for Founders gives you the edge you need to win in 2025
1:00:58
and beyond. Hit AI for founders.co to
1:01:02
subscribe to the newsletter, tune into the podcast, and grab
1:01:05
your spot in the next workshop, and it's all free.
1:01:09
Build smarter, move faster, think bolder with AI for
1:01:13
Founders. Now back to the show.