Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
CoinFund x Perle AMA
Recorded: Sept. 2, 2025 Duration: 0:46:40Space Recording
Full Transcription
Thank you. Thank you. Thank you. Thank you. Testing one, two, three. Thank you. . Thank you. Hello, hello. Testing, one, two, three.
Awesome.
All right, folks, we're going to get started here in just a second.
Testing, one, two, three.
Can you hear me?
Hello?
Yes, sir.
I hear you.
Great.
Hello, this is Ahmet.
Testing, can you guys hear me yes sir can you hear us perfect
all right we're almost ready then
all right well since it's it's time maybe we can just jump in. So hello, everybody. Welcome back to yet another CoinFund Twitter Spaces. My name is Jake Brookman. I'm CoinFund founder and CEO and managing partner. And today I'm joined by Evan Fung, who is our partner at CoinFund and Director of Research.
Evan, what's going on?
Jake and fellow listeners, welcome.
Really excited to be here and excited about today's conversation.
As some of you may know, a couple of weeks ago, our portfolio company, Pearl, announced its $9 million seed round led by Framework Ventures.
We at CoinFund are honored to have participated and continue to back Pearl founder Ahmed Rashad,
who is on the call and representing Pearl,
who is at the forefront of how human expertise
can be leveraged to train AI data.
So always an exciting topic.
Awesome, thanks, Evan.
We're excited to have him on today's Twitter spaces
to unpack the fundraise and everything about company.
Please welcome Ahmed Rashad, founder and CEO of Pearl.
Ahmed, are you there?
Hello, folks.
Yeah, thank you very much.
I'm very, very excited to be here and very excited about the future of this.
Okay, fantastic. So maybe for anybody who might not know you yet, maybe you could start with an introduction about yourself. And then what is Pearl?
Yeah, absolutely. So I'm the founder of Pearl. I started Pearl in 2024. And before that, I've done a whole lot of things. They might seem very disconnected, but they kind of are actually.
lot of things. They might seem very disconnected, but they kind of are actually. So I was an
offshore oil driller, and then I thought myself that I could and built a product and got acquired
and became a software guy 20 years ago. Did a whole bunch of things, really got into operations
research and optimizing complex systems, studied at MIT, and worked at companies like Amazon,
and worked at companies like Amazon and then Scale AI,
where I was leading all human data there.
So, and I took them from very early on to where they are.
And I noticed something that was missing
in the future of data.
And basically, Perl is the infrastructure layer
for high quality AI training data.
So we're solving a problem that I've lived through
through a lot of these places like these,
retail, Amazon, and with hundreds of customers.
Most of their labeling is either expensive
or really bad sometimes.
But what we're doing is basically a system
that can handle the super complex nuanced data
that modern AI needs
and what the future AI applications need
versus what was built already, which is a little bit behind.
And think about training models for autonomous vehicles,
for example, or medical diagnoses or frontage models.
You can just throw a generic,
you can't throw a generic label at it.
You need people who understand the domain
plus systems to maintain quality.
And we raised this year round because there's a massive gap opening up between what the AI companies need and what existing data providers can deliver.
Thanks, Ahmed.
Ahmed, maybe just to turn the screw one more time on the detail of that, maybe you could talk a little bit about what scale was doing for the AI industry and how big that was.
I know scale has been variously in the news recently, but just for anyone who might not be as close to the AI industry, what is training data and why does it require like a whole separate
business to provide it? Yeah, yeah, absolutely. So to build, to build a model, to build any model,
you need three main things. You need, well, researchers who write the algorithms and prepare
the models. You need compute to be able to process it. And you need a lot of good data.
If you don't have compute, or if you don't have the researchers, you're not invited to the table.
So once you have those two things, what really sets you apart is how good is your data?
Like I've seen it over and over again that weaker models with better data outperform better models or better built models with bad data.
from better models or better built models with bad data.
So, because you need very, very large numbers of people to do work consistently
across over a long period of time.
And that is a very, very difficult problem.
So that's why I joined scale.
When I started talking to scale, there were valued at 14 or $16 million. And anyway, the company grew very, very fast, actually, because once we unlock the human side of data preparation, that's what customers needed at the time. And it was great. And we unlocked it. So the company grew just recently, 49% of the company got a meta for $15 billion.
The total company valuation is roughly $29 billion now.
So that's what Sale did for a lot of the AI applications
you see today with autonomous vehicles.
But the future of AI applications requires significantly more specialization.
It's not about autonomous vehicles in boxes around.
It's about highly specialized stuff.
Legal.
Yeah.
Niche, Web3 and so on.
Got it. Yeah. Staying on this topic and maybe what the future of data services
and data labeling specifically look like,
how do you see kind of the push versus pull dynamic
in terms of Perl's product strategy?
And, you know, how typically do you go about designing
and deciding what to build next?
Is it a matter of what your White House customers need?
Is there a bit of, I guess,
prospective or anticipatory building
in terms of growing the company in a way
that can kind of skate to where the puck is going,
so to speak?
Like, how do you think about those two,
maybe, demands that are in tension with each other?
So the short answer is it's a combination of both. So there are cases, right? demands that are in tension with each other?
So the short answer is it's a combination of both. So there are cases where we get into,
for example, actually, we got that into a few use cases for niche languages and medical diagnoses.
And so we started researching the market there. It's like, oh, there's a huge demand. We understand where where the models are falling short because the data actually
helped our customers build models.
And the only way that is what they're building,
that you can provide them with the best data.
So there's this.
And it's where the about.
We also a bit of research.
And we start preparing those.
So on that note, I just want to clarify something else.
Like the whole data is the new oil thing is actually a little bit backwards.
Like in my mind, oil is a commodity.
Data isn't.
Quality data is becoming the biggest differentiator in AI development.
And we are moving past the scale at all cost era.
I mean, look at what happened with GPT-4 versus earlier models,
right? And GPT-5. Like the biggest improvements came from better training data and curation,
not just turning more compute at it. The marginal returns of adding more compute,
they're flattening out, but better data, that drives real performance gains.
Everyone you can think of, OpenAI, Anthropic, Google, they're all heavily invested in data
quality because that's where the next breakthroughs are coming. And that's where a lot of the push and pull is coming from.
So a lot of use cases are established today. Several use cases, people still don't know what
they're, they're still doing research and we are doing research alongside with them to try and fix
fix these or try to help them build the right pipelines for the right models.
these or try to help them build the right pipelines for the right models.
Yeah, that's really helpful, Ahmed.
Maybe we could talk a little bit about the different kinds of AI data that exists out
there.
And what I mean is, we hear a lot about, for example, on the one hand, data cleaning.
On the other hand, we hear about synthetic data and how we can actually produce data using you know llms themselves for training other llms um could you walk
us through a little bit about like you know what kind of data is most useful and um you
know which of those types is pearl supporting today.
Did we lose Ahmed?
Are you there, sir?
It looks like he's fixing some audio difficulties on his end.
We'll give him a second.
Yeah, I was hearing Ahmed go in and out a little bit. I don't know if you guys were hearing the same. ...
Are you there?
Otherwise, we can maybe take a slight detour while we're waiting for him to fix his audio.
And we certainly talk about, you know, Jake and I, in CoinFund's thesis on Perl, there
are a lot of exciting components.
You've heard some of Ahmed's background.
We also found that extremely compelling.
And I think one aspect of the thesis that is not fully maybe appreciated
as it's still a work in progress
is obviously there's strong alignment
between the kind of decentralized AI thesis overall
that Jake especially has kind of spoken about at length
on behalf of CoinFund and some of the product future
that Perl has also kind of expanded on. So maybe we could spend a little bit of time on that,
but maybe just checking back in on if Ahmed's coming back.
Yeah, it looks like he's restarting.
But I would just say to that point,
Ahmed kind of glazed over this a little bit.
But he was, as he said, he was leading all human data at scale.ai.
So he was building, you know, basically like these big supply chains of human workers globally
that, you know, at that point were still very much needed to create high quality outputs.
And that's a big job.
Ahmed, are you able to speak now? Are you back?
Yes. Can you guys hear me now?
Yes, we can. Yeah. So sorry. So just to get back to where we were, my question to you had been,
you know, can you talk a little bit about like the types of data, you know, human generated,
synthetic, et cetera, and what kind of, which of of those types is Pearl most excited about?
Yeah. So when you think about synthetic data and human-generated data, we specialize in human-
generated data. Ironically, we use a lot of AI to help the humans do better. And sometimes to
train humans before we get the customer data, we need to synthesize some data ourselves. So we actually use synthetic data to help build human
data. And companies that are generating synthetic data actually need some human data for their
training. So they are a little bit intermixed in the kitchen. But when it comes to actual
applications, each one, human and synthetic, has its own time and place where it's used and i don't think one
can completely replace the other yeah that makes sense so you know when you think of like scale as a
you know web 2 company and and pearl is a web 3 company but they're both working on this
you know human supply chain for data.
What sets the Web3 company apart? How do you guys think about that?
Yeah. So the human data part is very critical. So what sets us apart is two main things.
Human data is all about humans. So you have to get four things right. You have to hire the right
people. You have to give them the right tools, give them the right training, and give them the
right incentives. Web3 is a game changer when it comes to incentives and finding the right people.
So two things set us apart, operations and domain expertise. On operations, when I built
scales processes from the ground up,
so I basically know exactly where
the quality breaks down,
some hundreds of projects and most failure
happens in the handoffs between annotation,
review and delivery.
We designed our entire workflow
to eliminate those points of failure.
Then on domain expertise,
we focus on the complex stuff
that generic providers can't handle,
whether it's training for multimodal AI agents,
edge cases, computer vision, you want language understanding tasks.
We have people that actually understand what good looks like.
The better way to find and retain and incentivize and engage those folks is leveraging our Web3 chops,
which really sets us apart
because we're solving a problem for the customer.
That's what we do.
And we use the best technology available.
And the best technology in this case
happens to be a combination of AI,
traditional software engineering, and Web3.
Awesome. Evan, over to you.
Yeah, maybe just to talk a bit more, obviously, without kind of discussing some of the earlier customer conversations, but how are you thinking about the Lighthouse customers that you are targeting?
Or maybe from a vertical perspective, you gave some examples of, for
example, niche medical data that there's a need for.
Is the sales process more outbound, more inbound driven, or how are you kind of finding the
opportunities, even given, it sounds like there's more opportunity than ever, especially
post-scale getting absorbed into meta.
So it so be great
to kind of hear about some of the go-to-market especially because that's probably one of the
areas most mysterious to just your typical listener yeah yeah yeah so so i'll tell you this
i'll start with this most of our customers have tried cheaper alternatives first and they come to
us because they need surgical precision they come to us because they need surgical precision. They come
to us because they need the domain expertise. Whether it's traditional AI company or a crypto
protocol, the standard is the same for us. We will deliver upnotch quality and there's no compromise.
That for the inbound versus the outbound, it's actually a combination of both. But usually what we've seen so far from the outbound, we reach out and customers are like, oh, we deal with a lot of vendors.
And we've never gotten to a POC point, to the point of POC approval concept where we haven't won the contract.
We have several contracts and several stories where they're like, oh, we don't know you yet.
You're still a young company.
We have established players on our roster.
We'll give you a fraction of the contract and every single time within a couple of weeks they're like hey we cancel the other contracts we're giving you more volume so you
also got to think about you got to think about um with the with with the scale news what's happening
is scale was those there was the big gorilla in the room, and they were very, very successful.
And I'm very proud of the work we've done there.
But you need specialists for the next generation.
So they optimized for different priorities than what the market needs now.
And a few weeks ago, I was speaking at AI4, one of the largest AI conferences, and I was talking about building robust data pipelines, and this came up constantly.
What I'm seeing and what everyone at the conference was talking about is enterprises are getting pickier
about data quality and specialization
and the one vendor for everything approach is not working.
It's breaking down.
It's just breaking down, right?
And this is where we come in, where we are specialists.
We pick up verticals or areas where we know
we can deliver significant value and we nail those
and then we add more verticals and more
and more specialties. So you can see that a big theme, I saw that a big theme from AI4 was the
data pipeline robustness is becoming a make or break factor for AI and the best model architecture
in the world will not save you if your data quality is low. You're just dead in the water.
And that is consistent.
And this is where we come in.
Our customers don't see us as just another vendor.
They see us as a research partner.
And several times they've asked us, can you actually tell us how do you think we should best use our data, not just draw a box around something for us?
Got it. Amazing. Maybe one quick follow-up question, then we can get to the audiences.
But where does the concept of exclusivity enter into some of these conversations?
Do most customers want some exclusivity?
Is it time box?
Do they not really care?
It's more of a race condition.
They don't want to pay more for exclusivity if they're the first requesting something
and they don't mind if somebody else, another customer of yours, maybe wants to use the same
data later. How do you think about that as, I guess, a variable in some of these discussions?
Yeah. So exclusivity goes both ways. So there's exclusivity of the customer only working with us,
and there's exclusivity of us working with the customer. So we see it both ways with the bigger players. So
if we're talking about Google and OpenAI and Samsung and so on and so forth, they typically,
they are not exclusive in any way, right? They're like, we want to maintain options,
we want to maintain some diversity. But in some use cases where exclusivity comes in,
is really on the edge and the edge use cases or with startups and companies that are working on something and they have serious competition or they're working on something that no one else is working on.
In that case, they are happy to pay 10x so that we don't work with their competitors because they understand that this is their edge.
And we have to be very careful here because we're basically betting on who the winning horse is
and we're giving them the resources
to make them successful.
So in a lot of ways,
we actually help shape the outcome
of who's going to be the winner
in, let's say, medical transcription.
That's a good place to be and have a little bit of the Kingmaker power. Yeah, absolutely.
We have to make the right bets, but I think we've made the right bets so far.
Wonderful.
Yeah, maybe shifting gears then to taking any questions from listeners on the space right
now, feel free to request the mic and our host will pass it over to you
if you have any questions.
If not, we can certainly keep going.
All right, while we wait for that, for any audience questions, maybe we... Jake, did you want to jump in with anything else?
Yeah. I mean, there's so much to kind of cover here, but maybe just dooming out for a second, Ahmed, I know that you're super involved in the AI world for many years.
And you kind of mentioned GPT-5 as well.
How do you feel things are going in the AI space?
And is it as exciting as it was a year ago?
It's very exciting.
It's incredible.
The pace at which innovation is happening has definitely accelerated. And it's a flywheel because the better models and apps we develop, the more we can develop. It helps us accelerate actually the development.
observations here is that all the things that we were excited about, they are becoming a base layer
or infrastructure. So you think about GPT, four, five, six, whatever, like Gemini, Claude,
Sonnet, whatever, it doesn't matter, right? All of these are becoming infrastructure. And what I'm
really, really excited about are a couple of things. One is the next wave of applications,
which is not a generic LLM
that's just basically good enough for everything,
but it's highly specialized stuff.
It's applications that can analyze trading patterns
or can check for security links and so on and so forth.
So I think we'll see a lot of this new wave of innovation.
And I think the excitement should start shifting from the base layer to the application layer
or from the infrastructure layer to the application layer.
And basically, we're seeing more and more companies getting also very mature
and very excited about acquiring and cleaning up their data, which is becoming a very, very obvious trend at this point.
So when you see what people are excited about and what they're fighting about now, it's mostly how do I build something that's very, very specific and how do I get the data that enables me to do that?
something that's very, very specific.
And how do I get the data that enables me to do that?
One last thought here,
and this might sound a little controversial or provocative,
but you think about Meta's acquisition of scale.
Meta didn't acquire 49% of scale
for $14, $15 billion for nothing, right?
It was very obvious that others like Google
aren't going to like this, this but meta didn't care because
this basically meant that meta has the firepower to or has more firepower to get their data in shape
before everyone else which is which is a very interesting tactical thing that's how much meta
values just having a slight edge in the data space.
They're willing to pay $15 billion for it.
Yes.
I remember like the week after you mentioned that to me, there was a, there was like a headline that, you know, Google and Microsoft and all the big names that are like pulling out as customers.
Because they, yeah, I guess they care about the integrity and privacy of their data.
So that was a good call.
Yeah, it's incredibly expensive. And I don't think Mr. Zuckerberg is just spends money on a whim.
So I think he knows exactly what he was doing. Totally. What do you think is the probability that like a big, let's say, Web2 AI company or a big tech company, you know, retains like a Web3 data solution at some point?
Like if you were to put a number of years on that, what do you think that could happen?
So we're seeing we're seeing a couple of very, very interesting and very exciting patterns.
I've had customers come to me and ask me about, hey, we're very excited.
We heard you were doing Web3.
Tell me about what you're doing.
How are you leveraging that?
That's very exciting.
And we told them, and they're like, oh, can we also do this and that?
So we're seeing a lot of Web2 companies understanding, having a better understanding of Web3 and how it can give an edge.
And our customers, they don't care.
They are like, can you deliver this solution for us or not?
And when we show them that we can,
and they figure out that we're using Web3,
they're like, oh, great, we love this.
And our customers, I can't mention names,
but our customers are people that everyone uses every day, right?
And they're like, this is exciting, this is great.
Some of them even asked if they can invest,
if they can get some sort of discount for staking our tokens.
So when we launched one, so I was like, okay, great.
We'll talk about that.
Nice.
Was that more product?
Yeah, was that more kind of product driven in terms of the unlock and interest or was some of that maybe regulatory clarity emerging over the course of this year? I assume there's probably less perceived risk of corporate adoption generally, right? But I'm just curious if there was anything specific some of those folks were pointing to to get them more excited and interested? Yeah, that's a great question. From what I've seen, it's actually a combination of both.
So when we're talking to, like I can't say, let's say the Magnificent Senate, and they all have
Web3 divisions, and they all want to be involved, I think that having regulatory clarity and having a, and I'm not a policy expert by any means, but I think
having at least that potential and that comfort that there's going to be clarity on this is
exciting people to try and get ahead of the game. And then on top of it, they see the results and
the results are way better than a Web3 company alone can do or a web two company alone can do
and this combination is is delivering results for them so obviously they they they like that like
companies at the end of the day they want their product done with the least amount of effort
at the lowest total cost of ownership not price total cost of ownership and that's exactly what
we deliver like price wise we're not necessarily the lowest price, but we are by far, like by a
factor, probably we're probably a fifth of the total cost of ownership of the next best option.
Awesome.
For anyone just joining us, we're talking with Ahmed Ershad, founder and CEO of Pearl, one
of our portfolio companies doing AI data.
Audience, I know that you don't know as much
about ai data as uh ahmed does so you should have a lot of questions and you know if you if you'd
like to ask one please raise your hand and we'll bring you on stage um otherwise evan and i would
just keep going for for a bit um and i would i would like, Ahmed, would you, are you able, you know, you don't have to,
like, name customers or anything, but are you able to walk us through kind of a case study of
a customer and how you've engaged them and, like, what kind of work you've done with them
in terms of data labeling? Yeah, yeah, absolutely. Oh,
I get very excited about this. You know me, Jeff.
Yes, sir.
Actually, before I tell you a case, sir, like actually how I got into data was I was building
AI apps for Amazon in 2017. And it was, I realized very quickly, it's like, oh crap,
I have some of the best engineers in the world
and some of the best researchers in the world.
And I had a very large team and our models weren't doing better.
So what do we do?
What do we do?
So I started tinkering with the data and started playing with the data.
And I built my own data labeling operation and data collection operation.
Because very quickly, it was obvious that slightly better data is way, way more productive.
And that's how I got into this whole mess in 2017.
And one of the interesting use cases,
I can't mention the customer,
but I don't know if you guys noticed,
like when you get on a conference call
or like Zoom, Google Meet, whatever, Teams,
or when you're having a phone call with someone and there's background noise, right?
Like there's an alarm or something, or someone's talking in the background.
Like other people on the call can't hear them as much.
Or if you're clicking on the keyboard, some people can't hear them as much.
This is actually, part of it is because of some of the work we've done, because we had a customer
who was like, okay, we have these noises and we have all these background noises. We got to figure
out how do we remove them from the calls? So we literally had to go and collect and label and
transcribe a lot of data about people typing while on conference calls and
about someone sitting in a room and one room over, there's a blow dryer going on or a washing
machine so that you can isolate those sounds and literally make your call quality clearer
and better.
So literally, I was talking to someone, I was talking to a customer actually, and he
was telling me, oh, hold on a second, the fire alarm in the building just went off, and I couldn't hear a
thing. It was actually because of this type of work. So that's one of the use cases that I find
very, very interesting. It sounds simple, but actually designing the data pipeline itself,
like we had to figure out, okay, what sounds, how do we isolate them, what's the right environment
to collect them versus just collecting an order. So this was very, very exciting, how do we isolate them? What's the right environment to collect them versus just collecting an order?
So this was very, very exciting.
And I think a lot of people use it
and benefit from this every day.
Yeah, awesome.
That's a great use case,
removing noises from conference calls and other media.
Okay, I guess one more for me
and then audience, raise your hand or Evan jump in.
Like what, like how can, how can folks engage Pearl today?
Like if you're a, either if you're a business owner and you have a need for data, like what's the process with you guys?
for data, like what's the process with you guys?
Or if you're a kind of crypto world participant
and you love tokens,
what should you expect from Perl in the future?
Yeah, so if you need data
or if you need our help on the data
or developing your models,
just reach out to me
or reach out to sales at pearl.ai
or support at pearl.ai, like pretty much anyone at Pearl, or through our website, which is
pearl.ai.
Also on my Twitter account, which is AhmedZRashad.
And if you're someone who's excited about crypto, again, just reach out to us through our website or email me or send us a note through Twitter and we'll get it going.
Okay, maybe super quick.
I think something if you could speak to would be any open roles or partnerships, etc. that you're interested in or want to advertise there as well?
I mean, I'm sure some of the listeners now or through the playback would be interested in joining your team and exploring the journey kind of to the future of Perl together.
So if you want to shout out any open roles, that would be great too.
Absolutely.
Absolutely.
So we are constantly hiring for good engineers.
Like we don't stop.
I like good engineers. Like we don't stop. I like good engineers.
We're also hiring for sales positions
and we're hiring for several Web3 positions.
Several of those are on our website.
And even if a position isn't necessarily on the website,
just send us a note
because we have some rolling positions
that aren't fully announced,
but we're hiring big time
on both the Web3, the Web2, and the AI fronts.
So we're constantly hiring
and we're looking for top tier people
who are going to obsess about a problem
and deliver results.
Our motto, one of our motos is
we never split the difference,
meaning that it doesn't matter what happens.
We only deliver results to the customer.
If that doesn't, it's very binary.
You either delivered or you did it.
So we need people who are going to obsess about the problem, do some of the best work ever that's super impactful to the world around them, and have a lot of fun doing it.
It's terrific.
Yeah, also, sorry, one more plug.
Like, we have five core values, and if you go on our website,
you will see what our five core values are.
And one of them is literally, we never split the difference,
and, of course, ownership. To just give you a sense of the culture we have, which is, again, obsess over the customer.
And by the way, we don't say customer obsession because in our minds, and I made this very,
very clear.
I make this very, very clear to everyone.
We don't say customer obsession is a value because it's a table stick.
This is the core purpose of why we exist.
We solve problems for people and get paid for it.
Awesome. All right, folks. Any questions for Pearl?
all righty awesome i also wanted to send a shout out uh to jake and evan and the coin fund team
and the framework team and our investors uh bill ty and uh slava rubin and uh here and all our other great investors.
Thank you guys. This is going great.
Yeah, thanks.
Thanks, Ahmad.
Definitely really, really excited to follow the AI space
from the data perspective that you guys provide
and really, really important data
in the sense of, you know, market data,
like what kind of customers are out there
actually buying stuff and, you know,
what are they interested in?
And there's just a huge, you know,
diversity of companies looking at this stuff.
I can't wait until, you know,
Web3 starts also adding, you know,
value propositions that make it competitive with other solutions and as Web3 continues to go mainstream. I guess if there's no questions from the audience,
I would just maybe hand it over to you in case you have any parting thoughts for folks or what should people look forward to with Perl in the
future? And then maybe we could wrap. Yeah, absolutely. So a parting thought,
the future of the next generation of AI applications are not going to be generic.
They're going to be very specific.
To build specific applications, you need specialized services and you need specialized data.
The next big fight, or at least several big fights, are going to be about who has the data.
has the data.
Web3 is a critical component to be able to secure that data
consistently over prolonged periods of time.
And we're going to have some pretty exciting announcements
and launches in the next few weeks.
So please stay tuned.
And more big announcements coming.
Thanks, Ahmed. It looks like we do have at least one question,
Laibad, if you want to jump in.
Thank you for raising your hand.
Hey, hi guys. Great AMA.
I just have one question.
At Pearl, do contributors really contribute their own data,
or they just annotate or just label it?
That's a great question.
So it's a combination of both.
So in a lot of use cases, we have already collected the data
and the contributors, they help refine the data and make it better.
In a lot of cases, in other cases, we need to collect the data in the first place
so the contributors actually provide their own data.
Sometimes the person who contributes the data is not the person who actually cleans it up because sometimes you need different skill sets for collecting versus actually cleaning up.
And to answer your question, it's both and sometimes it's different variations of them.
So sometimes the contributor collects the data. Sometimes the contributor collects the data,
sometimes the contributor cleans the data,
sometimes the contributor does both,
and sometimes someone does this and someone does that.
A lot of times actually multiple people contribute to
the same piece of data because we need
multiple inputs from multiple people.
All right. I get it. Just one last question.
When you say high-quality data, what kind of data do you see most of the time? Is it into the healthcare, financial, or just human data or human patterns like you see on the web? What exactly kind of data are you guys working with?
A lot of companies are working on very, very exciting stuff.
Because we specialize, when we take on a use case, we go really deep on it.
So for us right now, we're working on language interpretation tasks.
We're working on medical.
We're working on legal.
And we're working on Web3 applications.
And those are the ones we dug deep.
We plan on adding more and more verticals in the future and
more and more use cases in the future.
So the answer is across the entire market, like I see data happening everywhere.
Robotics is going to be big, for example, it's one of the places where I'm placing a
few bets on.
And medical is going to be big.
Taking existing applications and translating them into other popular languages around the
world where they function natively is going to be big.
Dialects is big.
But also, that's the overall market.
But our, my view, is a little bit distorted because we don't take on everything.
The one vendor doing everything model is gone.
What we do is we take specific usepaces and we dive deeper so my view is a little bit
biased right
maybe 20% of my world
is just medical data
yeah right okay I got it
thank you so much for that guys
I actually know how to design prosthetic
dental prosthetics now
alright
thank you so much, man.
Awesome.
Thank you, guys.
All righty.
Well, thank you so much.
It was a pleasure having you on, Ahmed.
So check out Perl.ai, and we'll be looking forward to what we're doing in the future with AI data and Web3.
Thanks, listeners, and thanks, Ahmed.
Congratulations again, Perl, for the fundraise announcement,
and obviously we continue to be really excited by what you and the team are building.
Thank you, guys.
Stay tuned, everyone.
Pleasure.
All right. Thanks, everyone, for joining.
Take care.
