Case study artifact
PawStyle by MV Clinic ↗
The recommendation engine.
Live demo of the diagnostic recommendation engine built for an independent veterinary practice. Generate a synthetic patient case, run it through the engine, and see the ranked diagnostics a clinician would receive. With reasoning, priority, and projected revenue per item.
Stored in this browser tab only. Cleared when you close the page. Get a key at console.anthropic.com/settings/keys. The demo costs approximately $0.02 per case.
1. Generate a patient case
Generate a synthetic case from the persona library, or hand-tune the fields below. Twelve realistic archetypes are built in.
2. Engine output
Awaiting case input. Generate or enter a patient profile on the left, then click Run engine.
About the demo: This runs Claude (
claude-sonnet-4-20250514) against the Anthropic Messages API directly from your browser, with a structured prompt that instructs the model to act as a clinical decision-support engine aligned with AAHA preventive-care guidelines. Output is illustrative. It demonstrates the shape of the production engine's recommendations (ranked diagnostics with reasoning and revenue projections), not definitive clinical guidance. A production deployment would run on the clinic's own patient dataset, integrate directly with the PMS, and be validated against clinician acceptance data over time. This demo does not store patient data and cannot reach real medical records.
Demo vs. Production
What you're seeing vs. What we'd build.
This demo
- Synthetic patient cases generated from twelve archetypes
- Single API call per case. Returns structured JSON with diagnostics, reasoning, priority, and revenue estimates
- Runs entirely client-side with your API key
- Illustrative; no real patient data, no PMS connection
Production engine
- Ingests real patient records from the clinic's PMS
- Retrained on the clinic's own accepted/declined recommendation history (the compounding data advantage)
- Served from the clinic's own infrastructure with HIPAA-equivalent governance
- Clinician-in-the-loop: the engine recommends, the clinician decides and acts
- Measured against acceptance rate, revenue per visit, and diagnostic capture lift
Up next
Once you have 30 cases, patterns emerge.
Running one case proves the engine works. Running thirty proves the engine learns. The cohort dashboard runs k-means clustering and a decision-tree risk classifier in your browser. Live. On a 30-patient cohort.