Mantis Biotech Raises $7.4M to Build Physics-Based Digital Twins of Humans
Mantis Biotech, a New York startup, closed a $7.4 million seed round to expand its platform for building digital twins of the human body. Decibel VC led the round, with Y Combinator, Liquid 2, and angel investors participating.
What the Platform Does
The core idea: pull in data from motion capture cameras, biometric sensors, training logs, medical imaging, and textbooks, then run it through a physics engine to generate high-fidelity synthetic datasets.
An LLM-based system handles routing, validation, and synthesis of those disparate inputs before they hit the physics model. The output is a digital twin that simulates anatomy, physiology, and behavior.
Current Clients
At least one NBA team is already using the platform. The use case: model athlete performance and track jump pattern changes correlated with sleep and activity data over time. That's a concrete, measurable application.
Other stated targets include surgical robot training, medical issue simulation, and injury likelihood prediction based on performance metrics.
The Pharma Play
Mantis is targeting pharmaceutical labs and FDA trial researchers next. The pitch is delivering insights on patient treatment responses without needing the patients. That's the harder sell. Regulatory acceptance of synthetic data for drug trials varies significantly, and the FDA's current guidance on digital biomarkers in submissions is still evolving.
The athletic performance side has a clear feedback loop. A coach can verify whether the twin's injury predictions matched reality. Pharma trials have longer feedback loops and higher stakes.
The Data Problem This Solves
Medical AI consistently runs into the same wall: labeled, high-quality human data is scarce, expensive, and privacy-restricted. Synthetic datasets generated from physics-based models could reduce that dependency. Whether Mantis's synthetic data is accurate enough for clinical applications is the question the $7.4M is presumably meant to answer.
The NBA use case is a reasonable proof of concept. Sports teams have abundant, well-labeled performance data and immediate financial incentive to act on insights. It's a good starting point for validating a physics engine before moving into medicine.
Source: Techcrunch