Mantis Biotech, a New York-based startup, is building “digital twins” of the human body to tackle one of biomedical research’s biggest headaches: the lack of reliable data for rare diseases and edge cases. The company just closed a $7.4 million seed round to scale the effort, TechCrunch AI reports.
The core idea is straightforward. AI models in healthcare keep hitting the same wall: they work well with common conditions backed by large datasets, but fall apart when data is scarce. Rare diseases, unusual anatomical variations, underrepresented patient populations. Mantis wants to fill those gaps with synthetic data grounded in real physics.
How It Works
Mantis’ platform pulls from a wide mix of sources:
- Medical textbooks and imaging
- Motion capture cameras
- Biometric sensors
- Training logs and performance data
An LLM-based system routes, validates, and synthesizes these data streams. Then a physics engine renders high-fidelity models of human anatomy, physiology, and behavior. The result: predictive digital twins that can simulate how a real person might perform, recover, or respond to treatment.
The physics engine layer is what separates this from generic synthetic data generation. It grounds the output in realistic anatomical modeling, according to TechCrunch AI. Founder and CEO Georgia Witchel gave a concrete example: “If I asked you to do hand-pose estimation for someone who is missing a finger, it would be really, really hard, because there are no publicly available datasets.” With Mantis’ physics model, you just remove the finger and regenerate.
Where It’s Already Working
Mantis has found early traction in professional sports. One of its main clients is an NBA team, where the platform creates digital representations of athletes that track how their movements change over time relative to sleep, training load, and other factors.
Witchel described a scenario where a sports team could predict the likelihood of a specific NFL player developing an Achilles injury based on recent performance, diet, and activity duration. That’s the kind of predictive modeling the platform enables.
The Bigger Play
Sports is the entry point. The real ambition is healthcare.
Mantis is targeting several areas:
- Preventative healthcare for the general public
- Pharmaceutical labs running FDA trials
- Surgical robot training using simulated human bodies
- Rare disease research where patient data is ethically constrained
Witchel made a memorable pitch for how people should think about these virtual humans: “You know how when you see a three-year-old running around, and they have a Barbie, and they’re holding it by one leg and smashing it against a table? I want people to have that mindset with our digital twins.”
The point is serious underneath the humor. Real patient data comes with massive ethical and regulatory constraints. Synthetic digital twins could let researchers test aggressively without touching anyone’s privacy.
Funding and Next Steps
The $7.4 million seed round was led by Decibel VC, with Y Combinator, Liquid 2, and angel investors participating. The money goes toward hiring, marketing, and go-to-market efforts.
What stands out here is the timing. As AI models push deeper into healthcare, the synthetic data market is heating up fast. But most approaches generate data statistically. Mantis’ bet on physics-based simulation could give it an edge in producing outputs that actually behave like real human bodies, not just look like plausible data points on a spreadsheet.
The big question is whether pharmaceutical companies and hospital systems will trust synthetic twins enough to base real clinical decisions on them. That’s a regulatory and cultural hurdle as much as a technical one. But if Mantis can prove the models hold up in sports and early clinical work, the path to broader adoption gets a lot clearer.
Full details on the raise and platform are available in the original TechCrunch AI report.