Most CTOs don't start their career designing sonar systems for submarines. But Robert Stewart isn't most CTOs.
As Chief Technology Officer at Arbital Health, Robert leads a team building AI-powered infrastructure for one of healthcare's most complex challenges: value-based care contracts. In his conversation with Lucas Hendrich on CTO2CTO, he shares lessons from nuclear physics research, underwater acoustics, early machine learning deployments, and 27 hackathons that culminated in helping power vaccines.gov.
His path offers a masterclass in career evolution, unconventional team building, and finding meaning in hard technical problems.
When Subject Matter Experts Write Production Code
One of the most striking innovations at Arbital Health challenges conventional wisdom about engineering teams.
Rather than hiring data engineers and teaching them actuarial science (a field requiring years of study and multiple credentialing exams) Robert's team takes actuaries and teaches them to code. The result? A new role he calls "actuarial engineers."
"We thought maybe it's easier to go the opposite direction," Robert explains. "Can we get our actuaries to the point that they can write code?"
With modern AI coding assistants, this approach has proven surprisingly effective. Actuaries now write full-stack code that ships to production, paired with experienced software engineers who provide oversight, conduct code reviews, and use automated testing to maintain quality.
The advantage is clear: actuarial engineers understand the domain deeply. They know which contract provisions matter, how healthcare finance works, and what data quality issues to watch for. That expertise, combined with coding ability, creates features that would be difficult for pure software engineers to build—no matter how skilled they are.
AI as a Force Multiplier for Complex Domains
At Castlight Health, Robert's team deployed machine learning models in production as early as 2011, long before "AI" dominated tech headlines. The use case? Predicting healthcare prices and potential health conditions from claims data.
Working with 1.5 billion medical claims, his team built systems to reverse-engineer negotiated rates between payers and providers, identify patients at risk for chronic conditions, and guide employees toward better healthcare decisions. The work required custom tooling for monitoring, data quality checks, and managing model drift, infrastructure that cloud providers would eventually productize years later.
Today at Arbital Health, AI serves a different but equally critical function: making complex actuarial concepts accessible. Through natural language interfaces, clinicians and medical economists can query data, generate visualizations, and understand value-based care metrics without writing SQL or learning specialized analytics tools.
But there's a prerequisite: clean, well-structured data with rich context. "That works only if the data is well structured," Robert emphasizes, "and the context is well communicated to the AI."
This combination (subject matter experts, AI tooling, and rigorous data infrastructure) allows Arbital to scale actuarial insights that would otherwise require manual, time-intensive analysis for every contract.
The Right Way to Run a Hackathon
Over his career, Robert led or co-led 27 hackathons. His approach differs sharply from the typical "build from the backlog" events that many companies run.
First, he prioritized inclusivity. About half of hackathon participants came from outside engineering and product: customer support, legal, finance, and operations. Customer support teams, in particular, brought valuable insights because they understood pain points directly from users.
Second, he incentivized cross-functional teams by awarding points for diversity. Projects with members from multiple departments received higher scores, encouraging collaboration across organizational boundaries.
Third, he made the stakes meaningful. Winning teams chose charities to receive company donations, creating motivation beyond swag and gift cards.
When COVID-19 hit and the company went fully remote, Robert's CEO asked him to run a hackathon immediately. Three weeks later, 150 people participated out of a company of 500, and produced 11 features that shipped to production.
Some of that work eventually powered the search backend for vaccines.gov, allowing Americans to find COVID-19 vaccination sites. "You never know where some of these things are going to lead," Robert reflects, "and how they're gonna transform your business."
His advice for others? Don't use hackathons to clear the backlog. Focus on new ideas. Give people themes but not overly rigid constraints. Make it accessible to non-engineers. And if possible, run the first one in person—remote hackathons work, but they're harder to facilitate for teams just getting started.
Healthcare's Interoperability Problem
One of the more sobering parts of the conversation centers on why healthcare data remains so fragmented despite decades of digitization efforts.
Unlike telecommunications, where incompatibility would destroy the product, healthcare systems evolved with limited financial incentive to share information. Legacy platforms built in the 1970s and 80s still run claims adjudication for major payers. Electronic health record (EHR) systems are highly distributed across thousands of physician groups, each with different configurations and semantic interpretations of the same data fields.
Even well-intentioned regulations like Meaningful Use, which incentivized EHR adoption, couldn't fully solve the semantic interoperability problem. Data might technically be exchangeable, but its meaning can vary significantly depending on how individual practices configure their systems.
Robert describes the challenge with characteristic clarity: "There are so many individual physician groups, and they have their own electronic health record systems. There's a long tail of EHR vendors. The way they actually get configured and the semantics of what the data means can be interpreted differently."
Add privacy concerns, contractual restrictions, and the complexity of coordinating changes across a fragmented ecosystem, and the problem becomes as much organizational and economic as it is technical.
Yet this is precisely where Arbital Health operates—building systems that can ingest, normalize, and reconcile data from disparate sources while maintaining transparency and trust.
Mentorship, Mission, and Meaning
When asked about influences on his leadership approach, Robert highlights a mentor from early in his career, someone who understood not just engineering, but finance, marketing, and operations.
"He wanted to get involved in all aspects of the company," Robert recalls. "That created this interest on my side of not just being focused on my engineering work, but trying to understand marketing, trying to understand finance."
But beyond technical breadth, his mentor demonstrated something more important: empathy. He invested time understanding what motivated each individual contributor, what kind of work energized them, and how they learned best.
Robert also credits a book: Things That Make Us Smart by Donald Norman, for shifting his perspective. The core lesson? People learn differently. Some thrive on deep practice and repetition. Others want variety and novelty every day. Great leaders adapt to those differences rather than assuming everyone shares their preferences.
That philosophy shapes how Robert thinks about talent, culture, and mission. Startups are hard. The work can be grueling. Financial outcomes are never guaranteed.
What sustains teams through difficulty, he argues, is clarity of purpose. "You want to be in a role where what you're doing matters," he says. "The people that you're with, you like. And then you get the added bonus that it's a promising idea that can really take off."
In healthcare, that mission is clear. The U.S. system is expensive, complex, and often inefficient. Value-based care represents a shift from volume to outcomes, but only if the operational burden can be reduced.
Arbital Health's work (removing friction, enabling transparency, and empowering better decision-making) sits at the center of that transformation. For Robert, it's not just a technical problem. It's the kind of challenge worth solving.
Final Thoughts
Robert Stewart's career illustrates a broader truth: the best technology leaders don't just manage systems: they understand people, context, and mission.
From underwater acoustics to predictive healthcare models to empowering actuaries to code, his path demonstrates adaptability, intellectual curiosity, and a willingness to challenge conventional thinking.
Most importantly, it shows that solving hard problems in complex domains requires more than technical skill. It requires humility, collaboration, and a commitment to work that genuinely matters.
Listen to the full episode of CTO2CTO to hear Robert's insights on AI in healthcare, building unconventional teams, and leading with purpose.
.png)
