Eight weeks. That's all it took for Harbor—a San Diego software startup with two founders, a handful of customers, and $15,600 in monthly recurring revenue—to go from launching an AI data tool to signing a seven-figure contract to execute an entire clinical trial.
Most startups in this space spend years perfecting their software before anyone trusts them with actual patient data. Harbor went the other direction, and fast.
The company announced the deal in late May: a $1.93 million, three-year contract with a medical device firm to run a full clinical study from start to finish. Not just provide software. Not just consult. Actually operate the trial. At the time, according to the company's own metrics, Harbor had seven live trials and was pulling in $15,600 a month in recurring software revenue. The contract was worth roughly ten times what they'd booked for the entire year.
It's the kind of pivot that raises eyebrows—or should. Either Harbor's founders talked their way into something they're not quite ready for, or they've figured out a shortcut that the rest of the $82 billion clinical research industry has somehow missed.
Software Was Just the Opening Act
When Harbor emerged from Y Combinator's Spring batch, the pitch was straightforward enough: an AI-powered system for managing clinical trial data. The founders unveiled three automation tools with names that betrayed their Silicon Valley pedigree. Magic Build, which purportedly generates case report forms from protocol documents in under five minutes. Magic Capture, designed to pull data from source documents and slash manual entry by 90%. Magic Monitor, a remote system that scores data points by confidence level.
Standard startup fare for 2026—take a tedious manual process, throw AI at it, promise dramatic efficiency gains.
Two months later, the positioning had changed entirely. Harbor was now calling itself an "AI-native CRO," angling for a slice of the full-service clinical research organization market rather than just the comparatively modest $1.5 billion electronic data capture software segment. The company's second launch post made the shift explicit: Harbor was now delivering "both software and services" and hunting for introductions to clinical operations leaders.
The timeline is aggressive even by the standards of Y Combinator companies, where moving fast and breaking things is practically doctrine. Harbor reported five live studies at launch in March. By late May, that number had climbed to seven, and they'd landed the CRO contract. Whether they can actually execute it remains an open question.
Who's Behind This, Anyway?
CEO Albert Cai comes with relevant credentials. He led regulatory strategy at Biolinq, shepherding the company through first-in-human studies to FDA De Novo authorization for a wearable biosensor. Before that, he covered diabetes technology at Close Concerns and studied biomedical engineering at Michigan. It's the resume of someone who understands how clinical trials work, at least from the regulatory and strategic side.
CTO Nathan H. Leung brings a different skill set—software development chops from Google and Ramp, plus early-stage startup experience as the first employee at Jupiter, another Y Combinator company. He studied math at UCLA and computer science at Michigan.
What they don't have, apparently, is a large team. Y Combinator's profile listed just two employees, though LinkedIn suggests the company may have grown to somewhere between two and ten people. That's an awfully thin bench for running an end-to-end clinical trial, which typically requires regulatory specialists, clinical research associates, data managers, safety monitors, and site coordinators. Maybe Harbor plans to hire fast. Or maybe they're betting that AI can substitute for headcount in ways that would make traditional CROs nervous.
Early customers include Biolinq (Cai's former employer), NOVA Genomics, and University Hospitals. One client, MotilityCount ApS, reportedly launched a clinical study "in just one week" using Harbor's platform, according to a LinkedIn post from March. Speed, clearly, is part of the sales pitch.
The Market They're Chasing
Harbor isn't exactly entering virgin territory. The clinical trial software space is crowded and getting more so, with incumbents racing to slap "AI-powered" labels on everything from data capture to medical coding.
Medidata's Rave EDC platform still dominates—it ranked as the top choice for sponsors and CROs in a benchmarking report last year. Castor promises "Agentic AI from study build to lock." Merative says its AI medical coding cuts manual searches by more than half. Advarra launched something called "Braid," an AI engine, last September.
Everyone's talking about AI. Few, if any, are using it as a wedge to flip their business model from software vendor to full-service trial operator.
That's Harbor's bet. Rather than go head-to-head with entrenched EDC platforms, the company is using its software as a foundation for something bigger—a services business where the technology enables a leaner, faster version of what traditional CROs do with armies of people and decades-old processes.
The pricing structure reflects this hybrid approach: free for academic and nonprofit studies (a smart way to build references), $2,000 monthly for commercial work, and custom enterprise pricing. Harbor also checks the compliance boxes that matter in regulated industries—21 CFR Part 11, HIPAA, ICH-GCP, risk-based validation, immutable audit trails. They're a Google Cloud Partner and part of NVIDIA's Inception Program, credentials that signal they're serious about infrastructure and AI, respectively.
The Real Test Starts Now
Here's what the $1.93 million contract actually means: Harbor has convinced someone—a medical device company, no less—to trust them with a clinical trial despite having been in business for roughly a year and possessing a team you could fit around a kitchen table.
That's either impressive salesmanship or impressive naïveté on the part of the buyer. Perhaps both.
The company says it has another $187,000 in software contracts lined up for the next twelve months, on top of the CRO deal. Which means the services contract dwarfs their software business by an order of magnitude. It also means they'll need to hire, and quickly. Running a clinical trial isn't like shipping a software update. You can't just patch your way out of regulatory problems or site management failures. There are real patients involved, real data integrity requirements, real consequences for cutting corners.
If Harbor pulls it off—if they can deliver a successful trial while maintaining their software business—they'll have validated something interesting: a new category of AI-native service provider that treats software development and clinical operations as one integrated function rather than separate vendor relationships. In a market forecast to hit $129.8 billion by decade's end, built largely on legacy infrastructure and manual workflows, that would be meaningful.
But for every startup that executes a bold pivot successfully, there are a dozen that overextended and collapsed under the weight of their own ambition. Harbor has secured the hardest thing for any clinical trial startup to get: a paying customer willing to bet real money on an unproven company. What they do with that opportunity—and whether they can hire and scale fast enough to deliver on the promise—will determine if this is a feature or a company.
For now, they've got room to find out. And a ticking clock.
