Knit Health Lands $35 Million Led by NEA

Startups & Funding

Knit Health Raises $35 Million Series A to Deploy Med-BEAM Model

Knit Health raised a $35M Series A led by NEA to commercialize Med-BEAM, a healthcare world model trained on billions of clinical decisions and claimed up to 4x more accurate than frontier LLMs.

By
Nathan Brooks
Filed
Channel
Startups & Funding
Read
3 min read

Knit Health, a healthcare AI company, has closed a $35 million Series A led by New Enterprise Associates, with participation from Town Hall Ventures, SignalFire and existing investors. The money will speed commercial deployment of Med-BEAM, the company's foundational world model for healthcare, across health systems and other healthcare organizations.

The San Francisco-based company is building what it calls a foundational model to predict and improve patient care. Its founding team spans research and operations: Jonas Knecht serves as Co-CTO, Jonathan Kolstad as CEO, Maya Petersen as Chief Scientist, and Ted Robertson as COO.

Why another model for medicine?

The company's argument starts with a gap in the data. Healthcare's most important information is often not written down. It lives in the decisions clinicians make — the orders they place, the referrals they send, the diagnoses they reach, and what happens next. Large language models, trained mostly on text, struggle to understand not just what has happened to a patient, but where that patient is headed.

Med-BEAM is built to close that gap. The model was trained on billions of real healthcare decisions and hundreds of millions of longitudinal patient records, learning from clinical experience at scale. For each patient it constructs a living model, predicting the care that patient is likely to need and showing how different interventions could change the trajectory.

The ambition is an infrastructure layer rather than a single product. Knit positions Med-BEAM's clinical reasoning as a base for hospitals and health systems, life sciences companies, health technology companies, frontier AI labs and, eventually, consumers.

"Healthcare still operates too often on what amounts to chance: which doctor you see, which day you arrive, which hospital you enter," Kolstad said. "The best clinical judgment already exists in medicine. We're building the intelligence layer that makes that judgment available for every patient, everywhere. Med-BEAM can understand where a patient is headed, compare possible paths forward, and help identify which choices are more likely to lead to a better outcome."

What do the benchmark numbers show?

To demonstrate the approach, Knit ran a comparison between Med-BEAM and a suite of frontier-lab LLMs. The task: predict the clinical actions that would actually occur for hospitalized patients over a 24-hour period. No outcome was nominated in advance, and no model was told where to focus.

The results, as reported by the company:

  • Med-BEAM was up to 4x more accurate than the frontier models tested.
  • It ran 100x faster.
  • It cost 250x less per run.
  • It showed superior calibration, meaning its confidence levels more closely matched actual outcomes.

These are the company's own figures, measured on its chosen evaluation. Independent verification has not yet appeared. Still, a 250x cost gap is not a rounding error; if it holds under third-party scrutiny, it changes the economics of deploying clinical prediction at scale.

What does the round buy?

NEA's Blake Wu framed the investment as a bet on one model powering many workflows. "Med-BEAM represents a fundamentally new clinical intelligence infrastructure," Wu said. "One underlying model can power dozens of critical workflows from care management and patient navigation to clinical trial recruitment, specialist scheduling, hospital-at-home, and treatment-response prediction. That's what makes Knit Health's approach so powerful: the intelligence is foundational, while the applications are nearly limitless."

The list matters more than the adjective. Care management, trial recruitment and hospital-at-home are workflows where health systems already spend heavily on labor and where prediction quality translates directly into cost. A single model serving dozens of them is a platform claim, and platforms live or die on unit economics.

The $35 million will now test that claim against hospital procurement cycles, which are famously patient. If Med-BEAM's cost and accuracy advantages survive contact with real deployments, the foundational-model race in healthcare gains a serious contender.

Source: svdaily.com

Share this article:

More from Nathan Brooks

Nathan Brooks

Show full bio

Senior reporter covering people, launches and day-to-day practice in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.

1,031 articles

Related articles

« Previous article