Knit

Knit (Knit Health) develops what it describes as a world model purpose-built for healthcare: rather than being trained on language like large language models, the model is trained on the recorded actions of clinicians treating patients, along with EHR data. According to the company, the model learns how care unfolds over time, representing the range of possible patient pathways and the probabilities of subsequent events. Knit positions the result as a single, unified intelligence layer intended to replace collections of narrow point solutions, and states it has been trained on hundreds of millions of clinical actions.

The platform is delivered to health systems as APIs, with Knit engineers working alongside customer teams to design and deploy integrations for specific goals. Illustrative use cases published by the company include intelligent specialist routing (identifying patients who can be accelerated or redirected based on modeled care pathways), predictive patient flow (real-time analysis of patient journeys with continuously updated recommendations to unblock flow across a hospital), optimized care team allocation across physicians, advanced practice providers and virtual care, clinical variation insight derived from learning from top-performing clinicians, and proactive care recommendations generated by monitoring patient records in real time. Knit states the platform is designed for enterprise healthcare privacy and security, handling protected health information under HIPAA-aligned safeguards with encryption, role-based access and continuous monitoring, and that because the model represents uncertainty it supports assessment of its own predictions.

Business model

Knit sells its clinical intelligence platform to enterprise healthcare organizations, exposing model capabilities through APIs that its engineers configure and deploy alongside customer teams for specific health system goals.