Georgia Witchel spent years ascending frozen waterfalls. Now she's trying to scale something considerably more treacherous: the FDA approval process.
Her company, Mantis Biotech, announced in January that it had closed a $6.3 million seed round in late December 2025, with Decibel Partners leading the investment. The New York-based startup—part of Y Combinator's Winter 2026 batch and staffed entirely by three women—is building what Witchel describes as infrastructure software for medical device testing and regulatory submissions. Think of it as plumbing for an industry that's still, somewhat remarkably, running experiments on animals to predict human outcomes.
The timing might be everything here. Just one day before Mantis filed an SEC disclosure showing it had raised at least $4.825 million, the FDA released draft guidance suggesting that certain monoclonal antibody tests could skip the traditional six-month gauntlet of non-human primate toxicity studies. That came on the heels of the FDA Modernization Act 2.0, which in December 2022 quietly removed the blanket requirement for animal testing in drug applications—a statutory mandate that had stood for decades.
Witchel, who holds degrees in computer science from Harvey Mudd College and bioengineering at the master's level, argues the regulatory shift creates an infrastructure vacuum. "The industry doesn't have the systems to actually capitalize on this," she said in the company's announcement. Medical device makers, she contends, are drowning in fragmented data—sensor readings here, imaging results there, testing protocols scattered across incompatible formats.
Digital Twins Meet Regulatory Reality
Mantis positions itself as "Palantir for the biomedical industry," a comparison that's either audacious or apt depending on how much credibility you grant to buzzword-adjacent analogies. The platform attempts to weave together the entire medical device development pipeline: CAD design, computational simulation, bench testing, verification protocols, and ultimately the FDA's Q-Submission process, which lets companies get regulatory feedback before filing formal applications.
At its core, the system builds what Mantis calls "validated digital twins" of human anatomy and physiology. These aren't just pretty 3D models. The company claims its simulations predict biological responses and then validate those predictions against real-world outcomes—essentially creating a feedback loop that, in theory, makes the models more accurate over time.
Whether that vision translates to something regulators will actually trust remains an open question. The FDA has been receptive to computational alternatives, but "receptive" is a long way from "reliably approved." The agency still moves cautiously, and for good reason—medical devices quite literally keep people alive.
An Unconventional Path to Biotech
Witchel's background doesn't exactly scream "enterprise software founder." Before launching Mantis, she founded Louiza Labs, another YC-backed venture. And before that? Competitive ice climbing, where she competed at elite levels on frozen surfaces that would send most people straight to a therapist.
There's something almost too neat about that narrative arc—ice climber pivots to regulatory compliance software—but the throughline might be more prosaic than poetic. Both pursuits require meticulous risk assessment and an ability to navigate systems (frozen or bureaucratic) that punish mistakes severely.
She's not doing it alone. Mantis operates as a tight three-person team, all women, which in the male-dominated worlds of both biotech and infrastructure software stands out more than it probably should in 2026. Decibel partner Jessica Leão joined the board as part of the investment, alongside capital from StoryHouse Ventures, Y Combinator, Pioneer Fund, and Spot VC.
What Comes Next
The company says it'll use the funding to expand its data integrations, improve platform interpretability—a fancy way of saying "make the black box less opaque"—and scale beyond its initial focus on sports health and human performance applications. There are open roles for backend engineers, the kind of positions that suggest the team is still building core infrastructure rather than just polishing what already exists.
The bigger question hanging over Mantis isn't technical. It's market timing. If the FDA continues loosening requirements for animal testing, and if computational models gain widespread acceptance, then companies like Mantis could become essential infrastructure. But if regulators pump the brakes, or if the industry proves slower to adopt than Witchel expects, then the startup is building ahead of demand—a dangerous place to be, even with $6.3 million in the bank.
For now, Mantis is betting that the regulatory winds will keep shifting in its direction. After all, those draft guidelines didn't arrive by accident. They represent years of advocacy, scientific advancement, and mounting evidence that alternatives to animal models might actually work better for predicting human outcomes.
The ice, as it were, is starting to crack. Whether Mantis can climb fast enough to reach the summit before conditions change—well, that's the bet Decibel Partners and its co-investors are making.
