Oggie Nikolic - CEO & Founder at Lind

Welcome to another Founder Wisdom pod. We have Augi Nikolic with us today. He's a CEO and founder at Lind, located in Menlo Park, California. OG Canadian, studied at University of Waterloo, almost 15 years at Google. Very interesting guy. I'm gonna talk about machine learning, life sciences today, and Lind, Lind AI precisely. So Augi, welcome to the podcast. Tell us a bit more about yourself and what you're building.

Great. Um, hello everyone. Uh, hi Charles, thank you for having me. Uh, honored to be here. Um, Augie Nikolic, CEO, founder of Lynd. Um, Lynd is a clinical research operations, uh, platform. We partner with leading health systems across the US, Asia, and the globe to help accelerate and streamline, uh, clinical research, present more opportunities for patients, to receive more cutting-edge treatments and help remove a lot of the bottlenecks and barriers to getting better outcomes and hopefully better treatment results for their specific conditions. So pleasure to be here with you today.

Nice. Let's talk about your background. Straight out of Waterloo, Google recruited you. You worked at Google AdWords, Analytics, Checkout, You lasted 13 years there, which is incredible. Tell us about like that beginning of a career for you.

Yeah, so Google was an incredible start for any engineer, any data scientist, in terms of the opportunity to learn from some of the best and, uh, talent in the industry. I literally had a chance to work with folks who wrote our textbooks and back in school and who were formative industry leaders. And working at a large prominent tech company, you also get to learn what it means to build enterprise-grade systems, ones that are robust, that scale, that are performant, extensible. And so be it taking beyond sort of the academic learnings, you get to learn a lot more pragmatic engineering skills and what it means to run those. And I had the privilege to be part of some of those leading and prominent products at Google across the Google AdWords, the Google Assistant teams as well.

And how does an employer keep an employee theirs for 13 years? Like, how did you stay so long and what took you so long to launch your own company?

So I think I've always had the sort of entrepreneurial spirit. So even within Google, I was always looking for what are the new areas and new initiatives, new departments, new products that are coming up. And so I was often sort of had the opportunity as well as the blessed with the privilege to be able to join and partake in some of the earliest developments on, you when Google Assistant was coming up, for example, and helped establish that team, helped establish and grow the service there as well. And so while I was there for 13 years, it spanned many different segments of my career and many different projects and departments at different stages of their maturity. Post-Google, I think it finally came a point in time where I looked beyond. I was looking for more meaning. You know, I had collected a tremendous amount of expertise in the data science, ML, AI space while at Google, and I was looking for a new vertical, a new domain that I found personally appealing. And I found that in healthcare and bioinformatics, where we can help in the precision medicine space get more patients more access to this cutting-edge treatments. Hence the pivot to my— the second half of my career, the last decade or so.

And just before we, uh, shift to the entrepreneurial part of your life at Betteromics, you helped protect like Google's $100 billion ad network. Uh, so you were sort of, uh, playing against the fraudsters. How did that feel? What were your responsibilities back then? And I don't know, did it weigh heavy on you that you were the guy standing in between the bad guys and Google and saving them so much money?

Absolutely. That was an incredibly rewarding and stressful role to have. So I led the Google AdWords risk engine team that was in charge of building the fraud and abuse models to detect protect any types of policy violations, financial violations, and issues on Google's ad network. As part of that role, we made sure to protect the end user, the, um, uh, that were using Google's search results and clicking on Google's ads, that were of the highest quality and fidelity. It was unbelievably stressful, both internally. It was extremely stressful and demanding from some of the external stakeholders such as the advertisers, the regulatory industry as well. And it was also actually a field in AI/ML where the adversarial portion had not yet matured. And so typically when you're building the AI models, you're looking not to overfit on your current datasets and so on. We had actually realized we ran into a problem where the adversaries we were dealing with were constantly changing their behaviors. So we had to start developing these new and different types of modeling techniques where the models and the AI models we were deploying would be resilient and robust enough to withstand any of these kind of changes in patterns, uh, post-deployment. And so that was both intellectually a very rewarding time for me as well as and extremely sort of proud of the accomplishment that we reduced the fraud and abuse rates within Google's network by over 100-fold in my tenure.

Help me understand though, in layman's terms, what a Google Ads fraud looks like. I'm still not sure.

Yeah, so Google is built on trust for the end user. So when the user comes to the Google search results page, they search for their bank, they search for their travel agency or to pay their credit card bill or to pay for their medication or so on. Part of those results would be the organic search results and the other part would be the advertisements that are part of the paid ecosystem from advertisers. So imagine yourself, you go Google for the Bank of America, your, You get a bunch of links, you click on the Bank of America ad, which takes you to a fraudulent website that is masquerading and posing as the Bank of America account and steals your login and your credentials, and that fraudster is now able to empty your bank account. So those types of scams, those types of ways of deceiving the end user while leveraging search engines trust are pretty prevalent in the industry. And so our task was really to identify those types of issues and remove them as quickly as possible so that the end users are not exposed to those types of illegal activity.

What about on the AdSense front? Were you there as well, where there's AdSense members that would fraudulently Claim Google Payments?

Yeah, so that was another part of the— there's a sister team there collaborating with us on the— that was more on the actual AdSense types of policies and violations and abuse that you would get fictitious traffic, manufactured clicks, and other ways to scam both the network and the advertisers in this case and the Google's ecosystem and compromise kind of the quality of the overall. And so we encountered fraud and abuse from the advertising side of the two-sided marketplace, as well as from the traffic or the demand side and the supply side. And so we had policy and fraud and abuse detection mechanisms facing both of those sides and be able to look at the holistic picture and make sure that the quality of the Google's ad network is as robust as possible for the actually, the 99% of the good users that are operating within it.

Even in these early days, even before AlexNet and all of that, you were leveraging AI to catch the bad guys. You were leveraging AI to outsmart other human beings, correct?

That's right. So we had a portfolio of different models specializing in different types of policies, different content, different modalities, if it was text or image, if it was looking at the traffic and navigation. And so we were managing hundreds of different specialized models that were hyper-trained and focused on specific abuse patterns. And then we have to stand up the entire machinery and a lot of companies that exist today in terms of what it means to deploy models, monitor them over time, all of the infrastructure of what it means to have an enterprise-level AI systems that not only can version the models, you can train them, you can monitor them in real time, and then you can actually choose to upgrade them or, and get all of that, the full lifecycle of your model development built out. And so most of those did not exist 15, 20 years ago, so we had to go through our own learning curves. We have to go through our own kind of a build-out to build that machinery and tooling in-house.

And these behemoths of companies, if they don't take that aspect of them seriously, it can be the end of them, right? Is it like the PayPal analogy in which there was Max Levchin working on the actual algorithm to prevent fraud? Like, if this is not being taken care of, they're these billion-dollar behemoths don't exist, right?

Yeah, I think most of these marketplaces, two-sided networks, um, trust is really the currency that you're brokering. Um, whether it's you're ordering, uh, food from a delivery app, or it's an e-commerce site, or a search engine, or a job marketplace, you're brokering trust between two parties. Um, the more The more fraud, the more abuse, the more issues you have, the more that trust is eroded. And so while the different types of scams and abuses might differ depending on each of those different services, that trust is really paramount to be upheld. The value of that company is directly tied to the trust that the good actors in the system bestow on and entrust to that company. So it's one of these selfless, thankless roles. If you do your job, nobody notices. If you do your job poorly, it is an issue. If you do your job well, hopefully there are no fires to put out.

Yeah, and it's the lawsuits that these can cause that can go in multi-billion dollar damages. How did you shift it from this to life sciences?

Yeah, that's a fantastic question. So my father went through a prolonged cancer battle and I had a chance to be one of the primary caregivers there, see the journey of the patient and their experience across the medical systems, clinical trials, different procedures, diagnoses, the odyssey that it is, and the toll it takes on the patient., as well as on the family members. Um, and so I had, um, one of those moments in my career of, uh, wanting to do something of more, um, impact and meaning in that space. Uh, I was moved by the opportunity from the patient's perspective that we can and should do better. Uh, so I joined, uh, GRAIL, um, bio, uh, an early cancer detection company to help develop an early cell-free DNA blood-based diagnostic test where most patients' cancers could actually be addressed or treated or intervened with better outcomes, but simply go undiagnosed for far too long.

It's a pretty intriguing pivot. But I guess this became like front and center of your life. And what problems have you observed exactly in the patient journeys? Obviously, the patient journey is very painful and not optimized, right? Was that a massive red light to your brain? It was like— and what prompted you to actually build this solution? Because you must have not seen anything else in the market, right? If we talk about BedroMix in itself, right? Like, how was the landscape back then?

Yeah, so the, um, I think the medical field is, uh, significantly shifting towards precision medicine. There's a tremendous amount of information and data now becoming available through different modalities. So from the genomics and sequencing, uh, DNA transcriptomics omics, uh, all metabolomics and other analytes, uh, imaging as well. Um, so the, the fascinating part that I found, similar to, um, the— my early days in fraud, is you're looking for needle in the haystack. You're looking for the disease, uh, you're looking for the early signs, the early somatic mutations of that disease, and for the early detection tests. So it was actually pretty analogous to the modeling we were doing on the fraud and abuse work while I was at Google, where 99% of the time patients are healthy, 99% of the time the end user is not a fraudster. And so it's, you're really looking for the outliers. The other part of in medicine actually that I've been most excited about is because there's so many modalities now, so much information, You actually need thoughtful, well-developed AI to really suss and distill and make that information actionable for the clinicians, to really distill it down and help them analyze and interpret all of this. And we're getting now more and more into sophisticated clinical models that can offer more precise treatments, can model people's diseases more precisely. So instead of saying somebody has lung cancer, actually understanding that there's probably dozens of different histological subtypes and probably very unique disease profiles for each person and potentially personalized treatments. And so the field of medicine, I think, is sort of at the early stages of its renaissance, I think, in terms of the data-driven and informatics-driven wave. And so we're excited to be part of that.

And what led to the acquisition of Apple? Like, that's pretty rare. Um, yeah, like, was it the metrics? Was it, uh, were you just there at the right place, right time? Why did this make sense?

Yeah, that's a great question. So that was a part of our— the— my previous company at Betteromics, um, that was, uh, a multi-omic precision medicine platform where we were working with some of the leading hospital systems and cancer centers on devising sort of clinical decision support tooling to predict patients' outcomes, to help with residual— minimal residual disease monitoring and recommending treatments and therapies. I think the synergy with Apple is really be the one of understanding the patient's molecular profile of disease, understanding the patient's molecular profile of their biology, and then joining that with the wearables, with the lifestyle choices, with the environmental choices, and bringing together the wearables with the health record and with the omics becomes sort of a really even more powerful of a predictor of a dataset that's enriched. We can now understand and map patients' outcomes. We can monitor the patients remotely. We can understand and profile patients that are maybe at higher risk of recurrence, or higher risk of certain toxicities, progression of their disease, remote device monitoring to get early warning signs as well. And so there were a lot of synergies, I think, for that acquisition between the internal wearables, the Apple Watch, and some of the patient or consumer-facing devices paired with the health records and omics as well.

How did the acquisition feel? Was it stressful? 4 years, or, uh, yes, it's like looking at you, you always seem like pretty chill, pretty, uh, cerebral, but you also did like 13 years at Google, so already financially set. Like, how did the exit feel?

Um, well, and every startup, uh, is a roller coaster. Um, And I think it was a very rewarding one. It was a wonderful journey in terms of our accomplishments and what we were able to build, the team we built. We had a wonderful CEO in Angela Lai, my co-founder there. And it was also sort of bittersweet to see that a prominent leading technology company saw it as worthwhile while to make that acquisition. Uh, it was extremely rewarding, uh, to see that, um, there was value, then there was a lot of continuity and longevity of the ideas, the work, uh, and the accomplishments that we had to put forward. Uh, I think it was also bittersweet in that a certain chapter, uh, concluded, uh, but also excited for sort of the opportunities that a larger, uh, company with more resources resources can enable to take it further.

And signing with Apple, like, aren't they quite restrictive when they acquire you? They're like, you can't start a company in life sciences or blah blah blah, or did you circumvent that easily?

Yeah, so we, um, I ended up starting Lind, uh, just before the, uh, the acquisition, uh, had concluded. Uh, it's also in different space in the non-competing space with the prior work. So we were able to get off the ground without issues there, and we've been really focused and excited with our new trajectory at Lynd in partnering with leading health systems now.

Where did you get the insight for Lynd while you were working? Like, where did this stem from? Do you think the I don't know, research or clinical trials are really useful for saving people's life. I do think so to some degree. Where did you get your insight?

Yeah, that's a great question. So I had many insights, both for my father's journey as a patient and what it meant to look at treatment options, navigate the care, and understanding what both standard of care approved treatments exist, as well as what research options and clinical trials exist. I think it is a journey and an odyssey that far too many patients struggle with, and they're some often left in the dark on their own. And then similarly, from the research side, both at Betteromics, at GRAIL, as we were running our own studies or partnering with health systems on research studies, I've seen what it means to stand up the studies, to accrue and enroll patients to the studies and the operational burdens it puts on the health institutions to conduct the research behind the scenes in terms of identifying patients, all the logistics, all of the documentation needed to conduct that. And so I identified an opportunity that there had to be a better way to graduate the clinical trials from a lot of manual data wrangling and spreadsheets and PDFs into sort of a technology-enabled solutions that can really help the research teams on the ground streamline and elevate and automate their workflows.

So how do you find all these studies? How do you find all the potential people that would be part of these studies? Is that the bulk of the work? Like, tell me, and you're used to like double-side market places, right? So how did you crack the chicken neck problem here?

Yeah, so we, uh, we partner with, uh, health systems and some of the leading and largest health systems in, in U.S. and Asia, and we continue to find wonderful health system partners that are always looking for, uh, uh, to do more and to help their research staff, uh, do, do more with, uh, platforms like ours. We support the research staff with all of their research portfolios. So they'll— these health institutions often have hundreds, if not thousands, of research studies already open and active at their health systems. We also help facilitate and bring more study options and opportunities, so more pharma studies to those health system partners, as well as support them with the ones they already have on the ground. And so our perspective has been that every patient should be considered and screened for every study available, whether it's industry or sponsored study, if it's investigator-initiated, so they have the most shots on goal and the hopefully the best clinical outcome for whatever their specific disease might be.

Are you sometimes frustrated with this slowness in life sciences and the FDA and all the compliance stuff?

So, um, I think our approach has been to embrace, uh, the fact that in healthcare, PHI and patient privacy, the regulatory landscape, the validations are all necessary and, uh, required. So we've designed our system in a way that is very patient and health system friendly. It deploys into the health system. The patient's records never leave the health institution, so it's very privacy-preserving and respecting. And then we've embraced all of the regulatory frameworks and have been compliant and, and have had to go through all of the standard industry audits and checks and compliance accreditations at And we see that as a, as a part of being in the healthcare field, in the life sciences field, uh, an enterprise-grade platform has to, uh, not just meet the bar, it has to far exceed the bar that's been set in terms of the quality of the results, interpretability, the data use rights, uh, the privacy considerations, the security considerations as well.

And how do you make money with this? Because these pharma companies need to pay the applicants to each studies. You just take a percentage? How does that work?

Yeah, that's a great question. So our approach has been to really partner with health systems with a pass-through fee model. And the reason why that is, it is, is such that the platform supports all studies equally. We did not want to simply take pharma contracts and have commercially biased results for a few studies. We actually wanted to provide a holistic, complete research platform for the health systems that can help support all studies, whether they're publicly funded, grant-funded studies, if they're privately funded and industry-funded studies. Consortium studies, and so on. And so our model is really to partner with the health system on a set of pass-through fees to all of the studies within that portfolio so that they can all be supported equally and that clinicians can be presented with all screening options that are available to a patient and make the clinical— a complete and informed clinical decision for what's the best treatment option for them next. Not one that's commercially biased.

And what's the economics on that? Because, yeah, it's quite costly to get human in these clinical trials. What is it, like $1,500 a day times 7? So it's like $10K per participant, and then you get what, 5% of that? Like, is that the model?

So we have set fees depending on the specifics. The size and phase of the study and the indication and the disease it's in. So we, we provide and work with the health systems on a schedule of fees. But you're absolutely right, the, the costs and the investment is significant in clinical research and, and studies. For every day saved, for every patient found, and, and the time to last patient in is really critical, um, to get those accruals done as quickly as possible. And not just from the cost perspective, it's also for getting those, uh, therapies and those interventions to market faster, um, is, is also very important. And getting them validated through, through the rest of the regulatory framework and getting them out into the market. Um, and so we can help de-risk a lot of that. The health systems that we work with, we can help accelerate a lot of these studies simply because we can identify more patients. We can ensure quality of the patients that are identified and work with the research staff to really be able to do more research with the current staffing that they have on the ground.

Interesting. What about AI in vitro studies, like having an AI avatar of yourself? Can you see a future in which we don't have to do like clinical studies on humans? Because it seems from your perspective it helps some people to have access to that treatment, right? And potentially save them. But yeah, what if we can do research quicker and prove the efficiency of something super fast and deploy it like almost in real time on people?

Yeah, that's a fantastic question. I think, I think there'll be a graduation of different levels of AI and how sophisticated it can get to modeling the full human biology. I think our approach has been really to provide the transparent, intuitive, clinically interpretable tooling to help still facilitate and empower clinicians and research staff with current traditional clinical trials. I think there are certainly companies out there looking to model molecular twin and in silico model the disease and the specific pathways. I think we're still some ways away from being able to fully simulate and conduct these sort of in silico studies. I would never say never, but I think we're still years, if not decades, away from that simply because the biology is really complex. Complex, that even in oncology, the tumor microenvironments, the stromal tissues around the person's biology beyond just the immune response. And so understanding and being able to model all of those minutiae details is really difficult in addition to just targeting the disease itself. So I— our premise has been really to help empower the current machinery of research and clinical trials and how it's devised today. But I could definitely see a world where simulated trials might be available at some, some not too distant future as well.

Love it. Augie, thanks for being on today. Where can people find out more about you?

Yeah, Lindback Care is our website. We'd greatly appreciate anyone to give us a visit. If they're health systems, if they're sponsors in pharma, we're always here. We're looking for new partners to help accelerate clinical research.

I was— Charles and Augie,