Georgia Witchel's résumé doesn't read like typical Silicon Valley fare. World-record ice climber. Autonomous surgical robotics founder. Now: clinical trial data wrangler.
It's the last pivot that has her making the rounds in Y Combinator's latest cohort, pitching Mantis Medical as something pharma companies desperately need but don't quite know how to ask for—what she calls "Databricks for biomedical and clinical data." Whether the industry sees things the same way is the multi-million-dollar question.
Mantis emerged publicly in early 2026 with a straightforward thesis: clinical trials are drowning in fragmented, ungovernable data, and that chaos isn't just an inconvenience. It's an existential risk. When a high-stakes drug program can collapse under regulatory scrutiny because nobody can trace where a number came from—Witchel points to the Simufilam debacle as Exhibit A—the infrastructure problem stops being theoretical.
"Data quality and fragmentation delays trials," the company states on its launch page, the kind of understated phrasing that masks a more uncomfortable truth: sometimes it kills them outright.
The Expensive Mess Behind Every Drug Trial
Anyone who's worked inside a clinical trial knows the setup. Data streams in from electronic data capture systems, clinical trial management platforms, laboratory vendors, omics providers, and—because this is healthcare—plenty of legacy CSV files that someone's intern is still manually stitching together. Each source speaks its own language. None of them were designed to talk to each other.
When trial operators need to analyze results, build predictive models, or simply figure out what's happening across multiple studies, they start from scratch. Every time. Rebuilding pipelines, reconciling conflicting formats, chasing down provenance. It's the kind of grunt work that delays timelines, inflates budgets, and occasionally produces the sort of data integrity questions that make regulators very, very interested.
Mantis is betting that life sciences will pay to fix this at the infrastructure layer—before the chaos compounds. The platform pulls from those disconnected systems and generates what Witchel calls "canonical datasets" with full lineage tracking, versioning, and biological meaning baked in. Not just rows and columns, but datasets that understand what a clinical endpoint actually means, what CDISC standards require, how omics data should map to trial outcomes.
The pitch is that once data is standardized and preserved with lineage, it becomes reusable. Analytics workflows, machine learning projects, operational dashboards—they all draw from the same clean well instead of everyone building their own plumbing. Cross-study analyses become feasible. Early warning signals for trial execution risk become possible. AI models get fed inputs that won't embarrass you in front of the FDA.
The company's platform is designed with HIPAA-compliant workflows and security and governance controls for regulated environments, though no independent audits of those controls have surfaced publicly yet.
A Founder With Range (To Put It Mildly)
Witchel's path to clinical data infrastructure wasn't exactly linear. Before Mantis, she raised $5 million for autonomous surgical robotics and digital twin technology for surgery. Before that? Elite ice climbing. Multiple world records. The kind of vertical walls that require calculating risk with unusual precision.
She studied computer science at Harvey Mudd, earned a master's in bioengineering from the University of Washington, and somewhere along the way accumulated enough expertise in simulation and human-performance modeling to pivot into drug development infrastructure. The company's Y Combinator directory page still carries traces of an earlier positioning—"digital twin company that combines LLMs with high-fidelity physics simulations"—language that sits a bit awkwardly next to the current data-platform messaging.
Perhaps that's part of finding product-market fit in real time. Mantis launched publicly around February 2026 with a demo video and renewed outreach in early March, when Witchel posted on LinkedIn: "We're launching Mantis Medical!!" The team is listed at three people, though full roster details remain under wraps.
Timing and Competition
Mantis is hardly alone in seeing opportunity here. Just days before the company's public launch, Unlearn announced a $50 million Series C led by Altimeter Capital for its own clinical trial digital-twin technology—AI-generated control arms that let trials run with fewer participants. Different wedge, same underlying thesis: simulation and better data can de-risk drug development.
Dassault Systèmes has been pressing virtual-twin technology into personalized medicine through projects like its Living Heart initiative, blending physics-based modeling with AI for cardiovascular applications. Established players like AnyBody Technology and open-source tools like Stanford's OpenSim already serve researchers building musculoskeletal and physiological models.
The broader tailwind is real. Regulatory agencies are warming to digital-twin approaches in biopharma—slowly, cautiously, but warming nonetheless. A mid-2025 industry overview in Pharma's Almanac noted increasing FDA engagement around validation pathways for these technologies, a signal that the experimental phase may be giving way to something more durable.
Still, this is life sciences. The industry moves at the speed of clinical trial design committees, not software sprints.
What We Don't Know Yet
Mantis hasn't disclosed customers, active pilots, or formal partnerships. The launch materials explicitly ask for introductions to senior data leaders in pharma and biotech—the kind of request that suggests early customer acquisition is still very much in progress. A "Case Studies" page exists on the company's website but wasn't publicly accessible at the time of reporting.
Funding details are murkier than you'd expect for a YC company. Wellfound lists a $4 million seed round dated September 1, 2025. A January post on a smaller news outlet claimed Mantis raised $6.3 million led by Decibel Partners, with participation from StoryHouse Ventures, Y Combinator, Pioneer Fund, and Spot VC, though this figure hasn't been confirmed by major outlets or official company releases, leaving the actual round size uncertain.
Infrastructure Plays in Healthcare: Hard but Lucrative
The core bet Mantis is making isn't technically novel—it's a bet on adoption. Can Witchel convince clinical trial operators—pharma giants, scrappy biotechs, contract research organizations—to pay for a data layer that sits between their messy legacy systems and the analytics they need to run?
The Databricks comparison is deliberate. Modern data warehouses became essential infrastructure for tech companies not because they were technically dazzling but because they solved a universal pain point at scale. Mantis is wagering that life sciences faces a parallel inflection: AI adoption is accelerating, regulatory scrutiny around data integrity is intensifying, and infrastructure plays in healthcare are drawing serious venture dollars again.
Whether trial operators see their data problem through Witchel's lens, though—that's the unresolved question. She's already demonstrated she can build in deeply technical, highly regulated spaces. She's also shown a willingness to attack problems that require equal parts engineering rigor and tolerance for ambiguity.
Now comes the harder part: convincing an industry that prizes regulatory compliance above nearly everything else that it's time to rethink the plumbing. And doing it before the next Simufilam-style data integrity crisis makes the decision for them.
