Someone like Harshel Bahl was always going to end up in science. Both his parents are physicians; he studied physics at UCL, and his academic work in the United States deepened a long-held interest in building a next-generation consumer genomics company. What he didn’t expect was that the fastest way to get there would run straight through the heart of the AI industry.
In the years since, Bahl has worked across healthcare startups, AI labs, and venture, accumulating the kind of cross-domain experience that most operators spend a decade trying to piece together. Now serving as a Strategy Lead for RL Environments at Bespoke Labs, he acts as the rare example of someone who treated every role as a deliberate investment rather than a destination.
A Science Education With a Business Agenda
Growing up in the UK with two physician parents gave Bahl an early and unusually granular view of how healthcare systems function and the different ways in which they can break down. He pursued physics at UCL, chasing a larger, more encompassing curiosity about the computational tools beginning to reshape biology and how those could affect fields like medicine. That interest led him to Harvard, where he specialized in bioinformatics at Harvard Medical School and conducted research at the Dana-Farber Cancer Institute, one of the country’s most respected cancer institutions.
Over the course of his studies, Bahl became increasingly drawn to the idea of building a next-generation consumer genomics company, something that addressed the shortcomings he saw in existing offerings. Bioinformatics was the credential he pursued in direct service of that goal. “I always knew I wanted to build my own thing,” Bahl says.
He’d taught himself to code independently before his formal academic work ever required it, treating programming as a strategic tool beyond anything else. He has always seen himself as a business thinker who became technical in order to make better decisions, not an engineer who later discovered a taste for strategy.
Straight out of college, Bahl began picking up the experience necessary to accomplish this. He joined the founding team of BetterBrain as tech lead, working alongside the founder of the McKinsey Health Institute. The company, which applies AI to cognitive health, has since attracted mainstream attention, including being recently represented on the Joe Rogan Experience. Bahl was there at the beginning, before any of that validation existed.
What the experience gave him, beyond a front-row seat to an early-stage company, was a clear-eyed understanding of how slowly healthcare actually moves once clinical advice and regulation enter the equation. He recognized that committing to a multi-decade play in that industry required operational knowledge he hadn’t yet accumulated. So he made a deliberate decision to go acquire it somewhere else.
Inside Scale AI at Full Velocity
Bahl joined Scale AI as a Strategic Project Lead during what he describes as the company’s most intense period, the six months leading up to its landmark deal with Meta, when the organization was, in his words, “redlining.” He started in a project management role and worked his way into something even more consequential.
Because he was among the more technically capable leads, Bahl landed a spot on a rapid-response unit brought in to fix high-complexity projects for Scale’s largest accounts. One of the most significant contributions from that period was an adaptation of a methodology for automated model evaluation that he deployed across multiple projects simultaneously.
From there, he moved to a centralized team responsible for overseeing process and tooling across the company’s GenAI business unit. He was, in effect, writing the operational manual that other leads would rely on.
The environment reinforced a philosophy he has carried forward: that unreasonable pressure on timelines forces people to abandon existing processes entirely in favor of something better.
“If you really push people, almost unreasonably, in terms of ambition, people will find new ways of doing things,” he says. “Rather than incremental changes, they’ll find a complete step-function, a different way of doing something – Scale brought the best out of people.”
Applying His Skills at the Intersection of Healthcare Strategy and AI Research
At Redesign Health, Bahl joined as the youngest member of a fund that spans pre-seed investment, seed, and venture building, a deliberate move into an environment populated by people with far more experience than he had.
The contrast with Scale AI’s youthful, high-velocity culture was immediate and intentional. His day-to-day involved building enterprise AI agents and evaluating companies that could accelerate the fund’s global operations, work that demanded both technical execution and the kind of higher-level strategic thinking that pure engineering roles rarely afford.
“I think at Redesign I got more time to think about grand strategy and higher-level things, rather than just getting in the weeds,” he explains.
That altitude proved useful preparation for what came next.
Bahl now serves as Strategy Lead for RL Environments at Bespoke Labs, an applied AI research lab focused on building the training environments that allow frontier models to learn complex tasks reliably. The lab conducts research into data recipes and reinforcement learning to produce proprietary models, and its RL environments business works directly with research labs to ensure those models perform with consistency at scale.
Bahl’s role sits at the center of that work: managing client engagements, steering strategic decisions, and applying AI research to produce outcomes at the edge of what the field currently makes possible.
“RL environments will be and currently are the key to building AI that actually hits the diffusion exponential across the economy,” he says.
From London to the Frontier of Reinforcement Learning
Harshel Bahl has moved through physics labs, health startups, and two of AI’s most demanding companies not by following a conventional path but by identifying what each environment could teach him. As AI continues to dissolve the traditional boundaries between research, engineering, and strategy, Bahl has positioned himself across the full stack of operations that will determine whether frontier models actually reach the industries and institutions that need them most.