01 · 感知临床
Ambient clinical sensing
Environmental and low-burden sensing that understands clinical state while minimizing disruption to patients, staff, and workflow.
Yilinjie Medical AI Lab
At the boundary of medicine and intelligence.
We connect sensing, modeling, action, and verification into auditable systems under real clinical constraints.
Clinical firstStart with real workflows and consequential failure
Evidence boundKeep every claim connected to a verification path
Safe actionKnow when to act—and when to decline
Research pillars
From ambient signals to trustworthy action
Four research lines share one goal: medical AI that is more usable, verifiable, and bounded in the real world.
02 · 建模世界
Surgical world models
Verifiable representations of time, phases, events, and risk—without confusing video generation with clinical understanding.
03 · 安全行动
Embodied clinical systems
Sensing, reasoning, and controlled action designed around human oversight, failure safety, and rollback.
04 · 可信验证
Evidence assurance
Protect provenance, timelines, versions, and reasoning so medical evidence remains intact through automated workflows.
Operating boundary
Capability expands. Boundaries do not retreat.
We distinguish research prototypes, internal validation, and deliverable software. A demo is never presented as clinical evidence.
01
Traceable
Sources, versions, assumptions, and failures are recorded.
02
Able to refuse
Insufficient evidence or excessive risk hands control back to people.
03
Verifiable
Endpoints and error margins come before capability claims.
04
Designed to evolve
Software ships one entry at a time; research status never masquerades as product maturity.
Lab notes
Latest technical notes
Methods, boundaries, and engineering decisions as they take shape.