In a startup landscape where AI is often hyped up as a product moat, it pays to have the ability to “AI-bend” or not only innovate how it’s used but the very technology itself to make the end-products more developer-friendly, scalable, and applicable to a wide range of applications.
That’s what Indonesian all-in-one verification platform Verihubs CTO and co-founder Williem Williem brings to the table. He talks about that and more in this call, along with (1) his approach to building tech teams, (2) his PhD research that unlocked Verihubs’ liveness detection advantage, and (3) how AI is driving not only Verihubs’ product, but also the company’s scale and profitability trajectory.
We previously called his co-founder and Verihubs CEO Rick Fernando to talk more about the industry pain points, market opportunities, and business model
Highlights and Timestamps
- (01:30) Paulo introduces Williem Williem;
- (01:48) Mindset shift from academe to startup; “One of the points that I found that is really important when building the product is to talk to your customers. When doing research, usually I’m just the one that takes the directions.”
- (02:54) Applying PhD research into scalable startup products; “The knowledge from my research previously [proves] that we can estimate the depth information of the face from only 2D images, [and that] really helps me to build this technology inside Verihubs.”
- (04:13) Verihubs’ origins from Williem’s perspective; “[Rick] found that identity fraud is a really challenging problem, and when we tried to discuss solutions, I [realized] that it might be solved with my expertise, which is artificial intelligence.”
- (05:29) From improving technology to making it useful; “When I was a researcher, I was only building a product that focused on public data sets that we are using to evaluate each other. But when we built Verihubs, when we built the products, what we focus on is that problem that really needs to be solved.”
- (06:41) Focusing on simplicity for client user and developer experience; “The focus here is the simplicity, because what we have found in Indonesia, especially the company that offers API products or API services usually have terrible user experience or developer experience.”
- (07:51) The role of AI in product-market fit, scale and profitability for Verihubs; “The keyword here is to support [our clients] to have faster decision processes.”
- (10:39) Building generic solutions that still meet a diverse array of use cases; “Make the generic version after talking to the customers; do not make the generic version without talking to the customers.”
- (12:43) Williem’s Key Learning from Y Combinator; “It is important for all engineering people…to not only focus on how to solve the problem using the engineering terms, but how to solve it using the business perspective.”
- (14:06) Williem’s Approach to Hiring and Retaining Tech Talent in indonesia; “The first keyword is nurturing the talents and the second one is that we have to talk a lot before we hire them.”
- (16:47) Biggest gap in Indonesia’s tech talent capabilities; “What I found is that there is a gap between the education outcome and the industry needs…”
- (18:06) Future of Verification Tech in the next five years; “What I found is that in five years, [we might see more around] verification in the metaverse…One thing that I haven’t found yet is how a real person can be verified in the metaverse.”
- (19:43) Rapid Fire Round;
About our Guest
Williem Williem is the CTO and co-founder of Verihubs. Prior to Verihubs, he was an artificial intelligence and computer vision researcher. He received his PhD focused on Information and Communication Engineering from Inha University, where his research topics covered computer vision, computational photography, and image processing. He has been published in top journals and conferences in computer vision, and has done several industry projects including mobile augmented reality for SKT Telecom, image quality enhancement for smartphones, and parallel processing for linear algebra algorithms at Samsung Electronics.
From Academic Research to User Research
Paulo: Just a little bit on Williem before we move forward. So Williem has had a pretty interesting background as a founder. So he actually comes from academia. So he had his PhD from Inha university, when it comes to computer science, computer vision, information systems, all of that, and coming from academia he went straight into Verihubs.
So, I wanted to ask you about that, right? Like how is it transitioning from the academe, from the research perspective into the CTO role at Verihubs?
Williem: It’s really challenging actually, because the mindset that we have when I was a scientist or researcher is totally different from the mindset that I have when building this real product that is used by many people.
The transition was actually really tough for me because [it took] some time, like one and a half years before I could make a stable and reliable product that many of our clients can use. One of the points that I found that is really important when building the product is to talk to your customers. When doing research, usually I’m just the one that takes the directions.
Paulo: You don’t need to care about any customers or users.
Williem: Exactly. That’s the mindset. That’s the way that I used to do things. And when I’m building the products, one of the important parts is talking to the customers. I have to talk with lots of customers. From those insights from the customers, usually we built the product that they really love and they really want to use.
“One of the points that I found that is really important when building the product is to talk to your customers. When doing research, usually I’m just the one that takes the directions.”
Paulo: I’m actually curious, what were you researching when you were back in the academe? Maybe you can share a little bit about that.
Williem: [My] research is usually around depth estimation. Actually, we estimate the depth or distance between the camera and the objects only from the image. By using that information, we can usually make various applications such as augmented reality, virtual reality, or many other applications.
The correlation between my research and what we are doing in Verihubs is that when we are taking an image of a person you have to measure its likeness and the likeness can be measured by using the information from the 2D image. Sometimes we use depth cameras like the one on the iPhone to estimate the depth.
But the knowledge from my research previously [proves] that we can estimate the depth information of the face from only 2D images, [and that] really helps me to build this technology inside Verihubs.
“The knowledge from my research previously [proves] that we can estimate the depth information of the face from only 2D images, [and that] really helps me to build this technology inside Verihubs.”
Paulo: It’s great that your research actually directly applies to one of the earliest products of Verihubs, which is liveness detection, which you mentioned. And actually, it also applies perhaps to self-driving cars. It actually has a lot of different use cases which is really interesting and great to know.
I would also love to hear from your perspective, although Rick talked a bit about it in his podcast, but also love to hear from you how you met Rick, how Rick approached you and convinced you to leave research and then get into Verihubs.
Williem: Actually Rick and I came from the same circle. We are really close with Hendra and Hendoko Kwik from Fazz Financial and Modal Rakyat respectively. When I was trying to find a real problem that I want to solve, real problems when people can use using my expertise in AI, computer vision and et cetera, Rick came up to me with his personal problem, which is the identity fraud in his hometown, where’s he’s spent like around $500 [because of it] and he found that identity fraud is a really challenging problem, and when we tried to discuss solutions, I [realized] that it might be solved with my expertise, which is artificial intelligence. We discussed this several times and we finally built Verihubs.
“[Rick] found that identity fraud is a really challenging problem, and when we tried to discuss solutions, I [realized] that it might be solved with my expertise, which is artificial intelligence.”
“AI-bending” to build scalable verification solutions
Paulo: It’s a great confluence of your expertise and then the existence of a pain point that can be solved by your expertise, because usually they just use tech for the sake of using tech, but this time around it really actually can solve the issue when it comes to onboarding customers, KYC, all of that.
We’d love to know — you talked about how being in a startup has really given you more of a focus on the customer, building things for them, getting feedback, all of that. How has your work with Verihubs influenced the way that you see this specific field of computer vision and AI and machine learning? How has it changed the way that you look at this particular technology?
Williem: When I was a researcher, I was only building a product that focused on public data sets that we are using to evaluate each other. But when we built Verihubs, when we built the products, what we focus on is that problem that really needs to be solved.
The researcher only focuses on improving the technology and improving the performance; they don’t really focus on the user experience. They don’t really focus on how clients can integrate with their service or product, technology or algorithms.
And when building the apps it’s really changed my perspective on research; instead of focusing on something new that improves the performance of algorithms, I will prefer to do research that can really solve the problems of human beings.
“When I was a researcher, I was only building a product that focused on public data sets that we are using to evaluate each other. But when we built Verihubs, when we built the products, what we focus on is that problem that really needs to be solved.”
Paulo: It’s still research, but just a different kind. Leading from that, like from a tech perspective, how then do you approach building this infrastructure platform? How do you approach building this verification platform at Verihubs?
Williem: As we’re building the platform at Verihubs, we are trying to make it as easy as possible for our clients to integrate with us. And that’s why we are trying to simplify the technology we have such as biometric verification, liveness detections, and also the connections with the government databases.
We are trying to simplify the developer and user experience that our clients have. And the focus here is the simplicity, because what we have found in Indonesia, especially the company that offers API products or API services usually have terrible user experience or developer experience.
We are excited to build this simple and easy to use verification platform so that any people, and any person can integrate with us directly and verify their customers.
“The focus here is the simplicity, because what we have found in Indonesia, especially the company that offers API products or API services usually have terrible user experience or developer experience.”
Paulo: I really like that it’s also focused on developers as well. They’re the ones who are building various products that would need verification. And I think the keyword here is really simplicity.
So one thing I wanted to focus a little bit more on — because there are a lot of startups that say they use AI, and sometimes it’s thrown around a lot, so would love for you to explain from your perspective as CTO, how does AI really impact the bottom line for Verihubs and impact the end-users? And then how does it also enable Verihubs to scale really fast as a company versus other types of verification platforms and grow profitability as well?
Williem: I think one of the strong points is the AI in [our products], the artificial intelligence technology that we have built in-house and it really brings advantages for us because we know the technology, we know how we can improve the performance of the technology tailored to our customers. Different customers have differentiated technology.
And the [main] point they use our AI technology is that we can help them to make faster decisions using our AI technology, instead of replacing what they have done before we are trying to use the AI to support their decision process so that they make faster decisions.
For example, when you are doing the customer onboarding, we are trying to have like 100% automated customer onboarding, [but] there are steps that should be done first before we have a hundred percent of automation.
And what we have found right now is that our clients have [already] benefited from us by [Verihubs] helping them to make faster decisions on whether these kinds of people are verified or not. So the keyword here is to support [our clients] to have faster decision processes.
And then, for the scale and the growth of Verihubs, AI itself plays a really important role because, as you know, by using AI we can help Verihubs [internally] to [also] have faster decision processes. And then we can as fast as possible because everything is being done by AI and because the AI technology is built in-house so that margin that we have is really big, and then it can be multiplied with other clients.
For example, when you build AI for one client, then they can be applied to a hundred clients without much modifications. And it really helps for the companies to scale at some point.
“The keyword here is to support [our clients] to have faster decision processes.”
Paulo: A couple of points there, which I think are really interesting. One is that the keyword, as you mentioned, for the impact for the users, which are most of the time developers, is really the ability to support what they already have. So you’re not necessarily replacing what they already built, but it’s makes it as easy as possible to integrate solutions that you have.
And then two, the scalability is something that is easily replicable. And because you guys don’t just know how to use the AI, but also how to improve the AI itself, the very technology itself, again, given your background, adds a lot more advantage in terms of how you leverage it.
One question and maybe I’m sure a lot of our listeners would have also is, given that the customers that you have have a wide range, right, sometimes you have banks, sometimes you have fintech like Payfazz, and you have other use cases as well. How is the AI made scalable while still catering to these unique user journeys?
Williem: It’s one of the lessons that we have learned during building Verihubs, because sometimes clients want solutions tailored for them. And we are trying to build the general versions of the customer’s needs, because sometimes when customers request something, they don’t really know how to use it. They don’t really learn how to maximize the power of these AI technologies.
And instead of relying on the customer needs, we are trying to talk with the customers and give some kind of explanation for them that our generic versions of the solutions are the best one for them to integrate, because they don’t really need to spend more resources on user experience testing or something because the solution that we have has been tested, has been researched by our team to provide the maximum capability.
For example, I can share that our liveness detections at first really focus only on the technology. And we gave the power for our clients to choose what kind of customer onboarding, and we just give it to our clients first.
But the lesson learned that we found is that we could not really give bandwidth for them. Instead we focused on building our kind of “general version” and they can easily use our generic versions, and it is much more [effective] in realizing the customer’s needs from that perspective.
Paulo: I really liked what you said about focusing on the generic version, and then also making the generic version easy to work with.
Williem: I think the important keyword should be to make the generic version after talking to the customers; do not make the generic version without talking to the customers.
“Make the generic version after talking to the customers; do not make the generic version without talking to the customers.”
Developing the next generation of Indonesian engineers and AI talent
Paulo: Talk to the customers first as that informs the generic version. And speaking of building with customers would love to also get your thoughts on your Y Combinator experience. What was the best part of that program, and of your personal experience through it? And what did you learn from the accelerator?
Williem: Building product-focused tech is really important. And I’ve learned that from Y Combinator because as an engineer and as a researcher I easily just focus on the solutions instead of the products and during the journey of me and Rick building Verihubs and while Verihubs attended Y Combinator, one key point is understanding the customers and build the product based on that.
And that’s one lesson that I’ve tried to deliver to my engineering team. Do not only focus on the tech stack, no matter how much of an expert you are on the engineering part. But the best engineering team is [composed of] the engineering people who understand the business process as well.
It is important for all engineering people all around the world to not only focus on how to solve the problem using the engineering terms, but how to solve it using the business perspective, and they can build much better solutions and much better engineering products when they understand the business.
“It is important for all engineering people…to not only focus on how to solve the problem using the engineering terms, but how to solve it using the business perspective.”
Paulo: I really love that point. Engineering teams and tech teams should also be aware of the implications of what they’re doing for the business, and how it really impacts the end customer experience.
And speaking of building engineering teams that also are familiar with the business, I would also love to know what’s your approach to hiring engineering talent, because again, like in Indonesia, it is especially a much more competitive landscape in terms of getting tech talent. So maybe you can talk a little bit about that and share how you approach hiring top tech talent in Indonesia for Verihubs.
Williem: It’s really hard. It’s really difficult to hire the top tech talent here. And especially because there are lots of successful startups that are trying to hire them as well.
And as we mentioned before, we are trying to do early-detect high quality talents by accepting many student internships from the various universities in Indonesia, because usually in the university in Indonesia, you’d have a one year internship program. During the one year internship program, we can improve the quality of each internship student, and some of our internships students previously are now the key foundations in our engineering teams as well.
Paulo: So they end up joining the Verihubs team afterwards.
Williem: Yeah, and the procedure here is that we do not wait until one year after they finish internships and then offer them full-time employment, but when they give some contributions for us, we are happy to offer them full-time employment and promote them early.
The first keyword is nurturing the talents and the second one is that we have to talk a lot before we hire them. Important talent don’t really see the financial, even though it’s really important, but they really want to see the goal, the vision of the founders, and to deliver the right vision, to deliver the right goal to them, then we have to talk a lot with them.
And that’s the thing that I’ve done for the past few years, trying to ask them for a coffee to meet and discuss what kind of goals that they have, individual goals that they have, and what kind of goals that we have, and we try to match that. And when we are matched with the goal, usually they join up with Verihubs and become the key employees in Verihubs right now.
“The first keyword is nurturing the talents and the second one is that we have to talk a lot before we hire them.”
Paulo: Yeah. I think that’s really great, really getting in early, because again, if you can’t find that talent, you make the talent, you train them and really create that next generation of AI that as you mentioned, and I’m sure, especially for those who are interested in AI, I think it’s really valuable for them to work with someone like you who has already that research background and really knows what they’re doing when it comes to this type of technology. If anybody listening out there is interested in AI and wants to learn more from that perspective, we’ll leave a link to apply to Verihubs in the podcast description. So you can check that out.
And what do you see as the biggest gap at the moment in terms of tech capabilities in Indonesia?
Williem: I think that’s the outcome from the education system because there are lots of successful tech startups that are hiring important talents and high-quality talents all around Indonesia, but the sources of the tech talents are limited at the moment.
What I found is that there is a gap between the education outcome and the industry needs, and that is being solved by various startups, HR startups, bootcamps, etc. But I still feel that the gap is still big and we need to have more bootcamps or non-formal education for engineers to learn.
And it will be really important if those institutions that are trying to solve the gap are not profitable institutions, so they are focused on how we build the talent instead of making some profits, so that there are more people who have access to high-quality content, high-quality materials and we can reduce the gaps between the industrial needs and the educational outcomes.
“What I found is that there is a gap between the education outcome and the industry needs…”
Paulo: That’s a really great point. It’s still a pretty big gap, and you guys are your part of this and the AI side of things and, really tying it into your talent funnel for Verihubs. So that’s great.
Now moving back into verification would love to know what’s your view of verification, platforms and the evolution of this technology in the next five years. Maybe we can talk a little bit also about different use cases that you see. Rick and I talked about fintech a lot, but maybe there’s some other interesting stuff that you’re seeing in terms of the customers that you have, or the different use cases that you work with.
Williem: I found that verification is a really, really huge problem and is a huge space and most verification platforms [will] focus on the fintechs and banks. We are trying to focus on not only the fintechs, but also other verification needs such as employee verification, scholarship verification or medical verification.
There are lots of verification problems that still [need to be] resolved. What I found is that in five years, [we might see more around] verification in the metaverse. We know that the metaverse is really hyped right now; there are lots of metaverse platforms and virtual reality platforms that are growing very fast.
One thing that I haven’t found yet is how a real person of Williem, for example, can be verified in the metaverse. Is the Williem in the metaverse real the real one in the world? And that might be really interesting and challenging for verification platforms to solve.
“What I found is that in five years, [we might see more around] verification in the metaverse…One thing that I haven’t found yet is how a real person can be verified in the metaverse.”
Rapid Fire Round
What are the top 3 traits a startup CTO should have?
Williem: The first one is […]. Second is excellence. And the third one is a sense of ownership.
What digital technology/innovation or sector (apart from the tech you are working on) excites you the most today?
Williem: Yeah, I’m excited about mixed reality, how we can mix virtual reality with the real world. I think it’s really interesting.
What is the biggest thing you have learned from Rick thus far?
Williem: The people-oriented mindset, how we deal with the people, with the employees [whom] we have and try to help them to grow — I think that’s the most important thing that I’ve learned from Rick.
What is your favorite book / resource when it comes to learning about AI?
Williem: It might be boring, but I love to read research papers from like top publications in computer vision. I read the iEEE conference on computer vision and pattern recognition proceedings.
What is your biggest advice for any, you know, researchers out there or people who are in the academe who are considering getting into startup?
Williem: Try to understand the customers. Build a product that focuses on the customers. Do not bring [in] your research mindset when you are building your startup.
Favorite activity to de-stress?
Williem: Well, there are two things. The first one is [listening] to Korean ballad songs and then the second one is seeing my child smile. I love those songs from like Sung Si Kyung.