Particle.news
Download on the App Store

Nvidia to Supply Open Models for Abridge’s Doctor–Patient AI

Abridge will fine-tune Nvidia’s Nemotron models with de-identified clinical records to run the system on its own platform with a target of clinical deployment later this year.

Overview

  • The partnership was disclosed on June 11 and pairs Nvidia’s Nemotron open-model family with Abridge’s conversational health data to build a model tailored for doctor–patient visits.
  • Abridge plans to fine-tune the open models using de-identified clinical transcripts and summaries so the system can automate visit notes, generate patient summaries, and support clinical decision tasks.
  • The companies say the model will run exclusively inside Abridge’s platform, a choice they cite as allowing on-premise deployability and lower ongoing compute costs than many proprietary models.
  • Abridge already deploys ambient-listening transcription across major health systems — Emory Healthcare says it has rolled the platform out to more than 3,000 clinicians — and Nvidia is an existing investor in the startup.
  • The move follows other tech–health partnerships and signals rising competition in clinically focused AI; next steps to watch include clinical validation, privacy safeguards for the fine-tuning data, and the planned rollout before year-end.