Alterra Labs announced a $12.5 million seed round led by Oak HC/FT, with participation from Flare Capital Partners and Company Ventures, to build simulated medical environments for testing and training AI models.
Alterra Labs approach to training medical AI
The New York startup evaluates models on medical tasks to identify where they fail, then converts those failures into targeted training exercises. Doctors and scientists help design the tasks and judge the models’ decisions, providing feedback that developers can use to improve their systems.
In a typical simulated patient case, a model might be asked which tests to order while clinicians assess whether its choices are medically appropriate. Alterra then turns clinicians’ assessments into scoring criteria and rewards used in reinforcement learning to encourage models to repeat stronger decisions and avoid behaviors experts have flagged as unsafe. The company also builds environments for tasks involving medical imaging, clinical trials and telehealth.
Founder and CEO Akash Kapoor said models often learn to operate software while still struggling with the clinical judgment behind each action. “The models don’t know when not to do something for a patient, and how to deal with uncertainty,” he said. He added that models tend to falter when asked to follow a patient over time, consider large amounts of information, or make a series of decisions, and that they often fail to recognize when they lack enough information to act safely.
Researchers have documented similar shortcomings. Studies using PhysicianBench and HealthAdminBench found that the best-performing agent completed fewer than half of the tasks; some agents described the correct action in a clinical note but never carried it out in the electronic health record, while others lost information as they moved between software systems.
Vig Chandramouli, the Oak HC/FT partner who led the investment, said Oak saw a common pattern across healthcare AI companies: adapting models for medical work required more than data alone. “All of the money that went into ‘I need data’ is now evolving into ‘I need to understand how to successfully do work,'” he said. He argued that creating realistic simulations of the software clinicians and administrators use — such as electronic health records, insurer portals and billing systems — requires people who understand both clinical work and model behavior. “You can’t build this stuff without clinicians,” Chandramouli said. “And more specifically, clinicians who actually understand how AI works.”
Kapoor’s background bridges both worlds. He studied machine learning before turning to medicine, pursued healthcare-focused machine learning research at the National Institutes of Health and completed three years at Columbia Medical School. He started the company in 2025 while attending Columbia, initially under the name ChartR, and worked with healthcare providers to study how frontier models behaved in clinical settings. He left medical school in his final year to work on Alterra full time.
Kapoor’s interest in healthcare safety also stems from his own experience as a patient: as a junior at UCLA he discovered a two-year-old X-ray report in his patient portal warning of a potentially cancerous tumor in his leg and recommending an immediate oncology referral. No one had told him; he was diagnosed with osteosarcoma and underwent surgery.
Alterra says it works with several frontier labs, hospitals, medical societies and applied AI companies. According to Kapoor, frontier AI labs are its fastest-growing customer group, and the company generates revenue from its training environments and post-training services. Kapoor expects that improving clinical judgment in models will require repeated testing and training as models gain capabilities and as medical knowledge evolves. “It has a lot of knowledge,” he said. “It lacks a lot of judgment.”
