All insights
Markets

Berkeley spin-out Knit Health raises $35m to sell an AI world model to hospitals

Its Med-BEAM model, trained on 140 million patient records, routes referrals, with the first three go-lives planned for early 2027.

By World Nexus Group Editorial

2 min read1 source

Knit Health, a start-up spun out of the University of California, Berkeley last year, has raised a $35 million Series A to take its healthcare AI model to market, chief executive Jonathan Kolstad told Healthcare Dive in an interview published on 19 November 20261. The model, Med-BEAM, has been trained on 140 million patient records1.

What Knit is selling

Knit describes Med-BEAM as a world model of the health system rather than a language model. It learns from sequences of clinical decisions recorded in electronic medical records, and predicts where a patient should go next1. Kolstad, a health economist, frames the target as a matching problem: getting each patient to the right clinician at the right time1. He called it “the most complicated matching problem in the whole economy”1.

The first product sits in hospital scheduling, routing referrals so patients do not wait in the wrong specialist queue. Kolstad pointed to rheumatology, where some systems are booking 12 months out and, he said, between a third and a half of patients who arrive after that wait turn out not to have rheumatologic disease1.

Knit's commercial timetable, as Kolstad set it out:

  • three go-lives are in progress for early 2027, aimed at February or March1;
  • the company is in contracting and sales discussions with larger health systems, and reports interest from safety-net hospitals1;
  • it has started initial work with payers1.

One design partner began with a single use case and its internal team then proposed 97 further applications, Kolstad said1.

What it means for the market

For investors, the round adds to a small group of companies trying to build foundation models from medical record data rather than text. Kolstad named the competition directly: Mayo Clinic working with Microsoft, and Epic's Cosmos and Curiosity suites, which build sequence models on medical record data and which he called the closest to Knit1. His case for Knit rests on data scale and on tuning the model to each health system's own practice patterns1.

For health system buyers, the pitch is operational rather than clinical. Knit is starting with providers, Kolstad said, because a clinical AI needs doctors' trust, and its first use does not tell clinicians what to do1. Payers are a second market: Kolstad argued that long referral waits push patients out of network, which a routing tool could reduce1.

Specialist access sits at the centre of the roadmap for nearly every health system Knit meets, according to Kolstad1. That gives the company a clear first problem to sell against, in a field where Epic already offers something similar inside the record system many hospitals use.

Caveats

The account comes from an interview with the company's own chief executive, and the claims about data scale, uniqueness and customer demand are his1. The interview names no investors and gives no valuation, and no hospital has yet gone live, with the first deployments due in February or March 20271.

Sources

  1. 1Knit Health CEO on the AI startup building healthcare’s connective tissue · Healthcare Dive

This briefing summarises publicly available research and reporting for information only. It is not medical, investment or legal advice. Follow the sources listed for the primary research and reporting.