Large language models trained on massive datasets can accelerate genomics research, streamline clinical documentation, improve real-time diagnostics, support clinical decision-making, accelerate drug discovery, and even generate synthetic data to advance experiments.
But their promises to transform biomedical research often run into bottlenecks. Beyond the structured data that healthcare relies on, these models struggle in edge cases such as rare diseases and abnormal conditions where reliable and representative data is lacking.
New York-based Mantis Biotech claims to be developing a solution to fill this data availability gap. The company’s platform integrates disparate data sources to create a so-called “digital twin” of the human body, a synthetic dataset that can be used to build physically-based predictive models of anatomy, physiology, and behavior.
The company markets these digital twins for use in data aggregation and analysis. These digital twins could potentially be used to research and test new medical procedures, train surgical robots, and simulate and predict medical problems and behavioral patterns. For example, sports teams can predict the likelihood that a particular NFL player will develop an Achilles injury based on recent performance, training load, diet, and duration of activity, Mantis founder and CEO Georgia Witchell explained to TechCrunch in a recent interview.
To build these twins, Mantis’ platform first pulls data from a variety of sources, including textbooks, motion capture cameras, biometric sensors, training logs, and medical images. It then uses an LLM-based system to route, validate, and synthesize the various data streams, running all the information through a physics engine to create a high-fidelity rendering of that dataset. This can be used to train predictive models.
“We can take all these disparate data sources and turn them into predictive models of people’s performance. So any time you want to predict human performance, that’s a very good use case for our technology,” Witchell said.
The physics engine layer is key here, Witchell told TechCrunch. By building on the synthetic data generated and realistically modeling anatomical physics, the platform helps enhance the information available.
tech crunch event
San Francisco, California
|
October 13-15, 2026
“If you were to ask us to estimate the hand pose of someone who is missing a finger, that would be really difficult because there is no publicly available dataset that labels the hand positions of someone who is missing a finger. It’s really easy to generate that dataset because you just take a physical model, remove finger X and regenerate the model,” she said.
Because Mantis’ platform fills gaps in data sources, Witchell believes it has the potential to be widely used across the biomedical industry, where information about procedures and patients is often unstructured or siled across different sources and difficult to access. She highlighted edge cases and rare diseases where data is difficult to obtain because there are often ethical and regulatory constraints on including patient data in public datasets or using it to train AI models.
“You know what would happen if you saw a 3-year-old holding a Barbie doll and running around holding it in one leg and slamming it against a table? That’s the way we want people to think about our digital twins,” she said. “I think this opens people up to the idea that their humanity can be tested when they’re using virtual humans. I feel like right now people are operating with the exact opposite mindset, which makes perfect sense, because people’s privacy should be respected. In fact, I don’t think people’s data should be misused at all, especially when you’re using digital twins.”
So far, Mantis has found success in professional sports, perhaps because it needs to be modeled after high-achieving athletes. Witchell said one of the startup’s key customers is NBA teams.
“We’re creating a digital representation of an athlete. It basically shows how this athlete jumped, not just today, but every day for the past year. It also shows how their jumps have changed over time, compared to how much sleep they’ve had, compared to how many times they’ve raised their arms above their head,” she explained.
The startup recently raised $7.4 million in seed funding led by Decibel VC with participation from Y Combinator, several angel investors, and Liquid 2. The funds will be used for hiring, advertising, marketing and go-to-market functions.
Mantis’ next steps are to continue building the technology and eventually make the platform available to the public for preventive healthcare, Witchell said. The company also works to support pharmaceutical laboratories and researchers working on FDA clinical trials, aimed at providing insight into how patients are responding to treatments.
