Home/Works/MedLocal
Medical LLM · GENSHI AI

MedLocal

A medical LLM that runs inside the hospital. Discharge-summary drafting, AI voice charting in the clinic, referral-letter OCR. One defining trait: patient data never leaves the building. Nothing goes to the cloud — it's all self-contained on machines inside the hospital. It keeps working even with the LAN physically unplugged.

GENSHI AI Fully offline Discharge summaries Referral OCR
MedLocal
A medical LLM that lives on-prem — cutting the busywork around charts without letting data out.
0egress
Patient data never goes to the network
3features
Discharge summaries, voice charting, referral OCR
LANcut OK
Unplug it and it still runs
On-prem
Runs entirely inside the hospital
What

What it is

The thing that eats the most time in a clinical setting, I think, isn't the medicine — it's the paperwork. Discharge summaries, charts, referral letters. MedLocal points an LLM at exactly that, to cut the busywork. It drafts discharge summaries automatically, transcribes what's said in the clinic into a chart, and reads incoming referral letters by OCR. Less about flash, more about shaving one daily nuisance off at a time.

Why offline

Why fully offline

Charts and patient data can't leave the hospital. That's the single most important constraint in medicine. A design that ships everything to a big cloud model for convenience trips over it on the first step. So MedLocal is built to be self-contained on machines inside the hospital. Cut the network physically and it keeps running. That data never crosses the building's walls is guaranteed by the architecture itself. All the experience of training image, speech and language models from scratch at my previous job feeds directly into this "keep everything at hand" design.

It's a drafting aid, not a decision-maker. Generated summaries and charts are built on the assumption that a clinician reviews and finishes them. It isn't a tool for making diagnoses or final calls.

Related

Related

← Back to Works