{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/fd8e71b75eb4456ab13c9148e6caf6a4\" frameborder=\"0\" width=\"1280\" height=\"960\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":960,"width":1280,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":960,"thumbnail_width":1280,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/fd8e71b75eb4456ab13c9148e6caf6a4-5752a0e4bf3f139c.gif","duration":929.622,"title":"MedFlow AI  — Clinical Intelligence Platform | GenAI Bootcamp Portfolio | Apollo Sebastian ","description":"DISCLAIMER: MedFlow-AI is an educational demonstration of AI in clinical documentation ONLY. It does NOT provide real medical advice and is NOT a substitute for licensed clinical judgment. All AI-generated content requires clinician review before any clinical action is taken.\n\nMedFlow-AI v2 is a Clinical Intelligence Platform for healthcare providers — built as one of the three portfolio project for the Generative AI Bootcamp with instructor Matt Dancho.\n\nThe problem this solves: clinicians spend 2 to 3 hours every single day on documentation instead of patient care. At $150 per hour for a physician, that is over $37,000 per provider per year in lost productivity. MedFlow-AI gives that time back — with age-appropriate AI tools for pediatric, adult, and geriatric patients, and a role-specific AI assistant for 9 different healthcare roles.\n\nWhat Is Demonstrated in This Video:\n7 KPI cards from live SQL — 8 encounters, 3 Critical ESI 1-2, 4 Urgent ESI 3, 12 abnormal labs, 1 pediatric, 3 geriatric, 13 RAG articles\nPatient Encounters — 3 ages expanded: ENC-2847 (58F STEMI — Troponin HIGH, ST elevation CRITICAL), ENC-2853 (8M pediatric — Rapid Strep POSITIVE), ENC-2854 (82F geriatric — subdural hematoma 4mm CRITICAL, INR 2.8 HIGH on Warfarin)\nTriage System — ESI 1-5 color-coded chart with interactive assessment tool calling the assess_triage OpenAI function tool — pediatric and geriatric special population protocols\nClinical AI Agent — LangGraph multi-step loop: get_encounter_details (live SQL) → lookup_clinical_knowledge (RAG scoring 13 articles) → synthesized clinical response. Demonstrated with STEMI and Sepsis Hour-1 Bundle queries\nCare Plans — existing pre/post visit cardiac plans for ENC-2847 plus AI generation for any diagnosis and age\nProvider Portal — 9 role-specific AI assistants: MD (Clinical Decision Support, Prescribing, Differential, Orders) and RN (Head-to-Toe 8-system assessment, SBAR, Patient Education)\nAnalytics — 6 live SQL charts and custom SQL query interface on the patient database\nClinical KB — 13 evidence-based RAG articles searchable and expandable — the grounded knowledge base powering every AI response\nCourse Skills Demonstrated:\nC1 — LangGraph Agentic Workflow: multi-step agentic loop calling 7 OpenAI function tools in sequence, chaining results between steps\nC2 — RAG: 13-article clinical knowledge base with keyword scoring — lookup_clinical_knowledge injects top-matched articles as grounded context\nC3 — SQL / BI Analytics: live SQLite patient database (encounters, labs, care_plans, progress_notes, daily_census tables) — 7 KPI queries, 6 real-time charts, custom SQL interface\nC4 — OpenAI Function Tools: 7 structured tool definitions with JSON schemas — lookup_clinical_knowledge, get_encounter_details, generate_soap_note, assess_triage, generate_care_plan, generate_progress_note, get_department_analytics\nC5 — Cloud Deployment: live HTTPS URL at medflow-ai-r2.streamlit.app — all 8 tabs fully functional — publicly accessible worldwide\nLive App: https://medflow-ai.streamlit.app/\n\nGenerative AI Bootcamp | Instructor: Matt Dancho | Apollo Sebastian | May 2026"}