Today's newsletter features my conversation with Anil Vaitla, CTO of NOCD and Noto, on building behavioral health infrastructure for conditions the system keeps getting wrong, upcoming AI meetups and Chicago tech events this week, and open engineering roles across Chicago's fastest-moving companies

NOCD has treated more OCD patients than any other provider in the country. Now they're building the platform underneath the whole category. Anil Vaitla is the CTO of both NOCD and Noto, the infrastructure layer powering a new generation of specialty behavioral health brands.

We spoke about what it means to fix a system that keeps misunderstanding the conditions it's supposed to treat:

AV: There are always challenges in any business. A lack of challenges may mean we're not growing or not making as large an impact as we could. Facing difficult things is natural and taking small actions to chip away at the problem is one way to manage this issue.

Furthermore, maintaining a close understanding of "why" we do the work we do helps. This can be another team member reminding us of what accomplishing a problem unlocks, listening to our members' concerns and seeing how the impact of our work affects them. For example, many share their stories of success after their experience with NOCD. Living with OCD can be a significantly more challenging experience than the technical issues we run into day to day and seeing how we can help others overcome the most challenging thing in their life gives us a connection to our purpose of serving and supporting others.

AV: I feel very fortunate for my time spent at Hulu. There was an incredible one team culture of support and a depth of technical excellence. I had great mentors and they were very open to teaching and being asked questions. They gave me a lot of freedom to work across the entire stack from mobile applications and frontend to video streaming all the way to database infrastructure to datacenter setup and hardware design.

I spent a good deal of time on reliability engineering which forced a good understanding of the fundamentals of operating systems to databases to networking and I was able to work every day on a new system I hadn't seen before - being able to work under time pressure when systems are down is a stressful but incredible way to learn quickly. I think anyone can learn anything with enough motivation and curiosity, but the guidance of technically exceptional mentors and colleagues is invaluable and accelerates your own personal desire to become better than you thought you could be.

AV: Generally, we stick to the best practices in software development. We place a large degree of focus on validation, debugging, and security given we operate in the Healthcare space, however there are aspects that we keep in mind given the misunderstood nature of OCD. It's always been our opinion that each condition needs to be treated in its own way, and lumping all behavioral health into a general one size fits all solution doesn't work well, as clinical literature demonstrates. For OCD in specific, ERP is far more effective than general talk therapy, the latter of which is harmful and causes the severity of OCD to worsen.

We spend a lot of time with clinical leadership to inform product decisions in our electronic health records system that clinicians use every day. We build content in a way to make the most misunderstood subtypes appear early on. We also focus heavily on brand and user driven community where people can connect with others who also have OCD and find someone they can finally relate to. As we now also have a sibling company Rebound for PTSD, we focus on keeping its own differentiated offering in order to emphasize the differences in those we serve and the clinical approaches to treatment.

AV: In the early days when you're unsure about what to build or what the right solution looks like you always have the answer in front of you. You just need to spend more time talking to your members to remind you of what's the most important thing to focus on. We certainly have many stakeholders apart from individuals receiving ERP therapy, such as the providers we employ to deliver the care and our insurance partners, so when there are times when we are unsure of the best step forward, we can simply ask and clarify it. Many team members and providers at NOCD also have lived experience with OCD and they can help recalibrate if something should be built differently which has been a core part of our development culture.

AV: The hardest thing is ensuring the focus we paid to our first condition continues on to every subsequent one. Generally this is a problem of scale, quality of the first product is exceptional, but the second, third, fourth product may not be. Most of the infrastructure that was originally built for OCD has been modified to be shared to PTSD. We can now scale quality because we don't focus on common elements that are robust but focus on condition specific nuance like clinical measurements, training protocols, and community development from day 1 vs having to rebuild the foundations again.

ERP has been around for a long time since the 60's and 70's. Edna Foa begin to formalize it and do research on outcomes for it which made it the Gold Standard of clinical treatment for OCD. This demonstrates that the clinical aspect in theory has been solved. The bigger challenge that remains is fairly multifaceted: building awareness that OCD is misunderstood as a personality quirk vs a debilitating condition, training a limited number of clinicians in ERP, getting reimbursement for specialized care, and ensuring quality at scale. Fixing the operational issue means deeply understanding OCD subtypes the mainstream misses, working with primary care teams, and demonstrating that claims data was missing the OCD population and miscoded as general anxiety and depression, showing clear cost savings to insurers.

AV: There certainly is a rush to put AI everywhere, but there should be awareness on the importance of testing and its limitations and failure modes (memory decay, counterfactual reasoning, continual learning, guardrails with rigorous evaluation sets). Guardrails, and in particular, manually curated evaluations, are crucial to getting things right. Development should begin with a clear evaluation test set so that results from inference produce the correct results every time. AI is incredible technology but missing the limitations of it (even from how it's trained) should be more honestly discussed.

AV: Chicago has a great ecosystem of talent and you don't need hundreds of people, just a handful of very motivated people who believe in the shared mission. The biggest downside of Chicago may be the frigid winters, but I think that makes folks more resilient and instills in people the importance of perseverance.

For professionals working in AI and technology, the next step in your career is not always clear. You may be considering a master’s program, MBA, or PhD; exploring a career pivot; or looking to strengthen your skills without committing to a full degree.

The University of Chicago’s Graduate Student-at-Large (GSAL) and Graduate Student-at-Large: Business (GSALB) programs are designed for exactly this kind of exploration.

They offer a flexible way to take graduate-level courses, build new skills, and clarify your academic and professional goals before committing to a full degree program.

Program Benefits

Join us for an online information session on May 27 at 1:00PM CT to learn how GSAL and GSALB can support your academic and professional goals.

Register for the Info Session

Follow me on Twitter and Instagram