- qwen3-8b
- extract · classify · code · gate
- single node
- llama-8b-2stage
- clinician + patient docs
- pipeline-parallel × 2 AI PCs
- scheduling
- local slot inventory
- deterministic tool
healthcare
Clinical referral triage
A referral letter is read, triaged against criteria with cited IDs, coded against an allowed ICD-10 list, scheduled inside SLA, and answered with clinician + patient documents drafted live on the 2-stage chain. A safety gate then checks the work before anything would reach a patient. PHI never leaves the premises.
pre-warming the mesh (qwen + 2-stage chain)…
The case