John Stamford

MrProf guides adults through the official ASRS v1.1 6-question screening and a one-question-at-a-time adaptive interview, then assembles a clinician-ready briefing where every point links to the user's exact words — quotes, line numbers, denials, uncertainty, and gaps included. It organizes what you report so a clinician can review it. It never diagnoses, rules anything out, estimates probabilities, or assigns severity. How it works: deterministic code plans each question, selects objectives, scores screening (original 0–6, ≥4 positive), and assembles reports; Meta's Muse Spark model (via an OpenAI-compatible endpoint + Microsoft Agent Framework) phrases questions and extracts evidence — all server-side, MongoDB as the durable source of truth, Docker Compose + optional Cloudflare Tunnel for the live demo. Evaluation is synthetic-only and honest: 6 hand-authored SPIT-style personas (vague, denier, contradictor, incomplete recall, 2 positive screens) run through the real API — 12/12 complete, 238/238 quotes exact-substring, 0 banned-language hits, backend suite 109 passed. Six frozen demo reports ship on the /demo gallery so judges can trace vague answer → targeted follow-up → cited evidence in 60 seconds. Topics and evidence codes are drafts, not clinically validated — stated once, upfront.