I just open-sourced AE Sentinel, an AI-powered pharmacovigilance platform that automates adverse event monitoring in clinical trials.
🔗 GitHub: https://github.com/mgj10086/AEPD-Sentinel
What it does
| Module | Description |
|---|---|
| 🏷️ AE Coding | Auto MedDRA coding with keyword-based synonym matching |
| 📋 SAE Reports | CIOMS-I report generation (PDF/DOCX/JSON) |
| ⚠️ Deviation Detection | 7 built-in rules, auto-triggered from AE processing |
| 📊 Signal Mining | Organ-class aggregation, Fisher's exact test simulation |
| ✅ Compliance | GCP compliance auditing |
| 📚 Knowledge Base | RAG-powered semantic search (ChromaDB) |
| 🔐 Audit Trail | HMAC hash chain, tamper-evident |
| 🔔 Notifications | Real-time alerts for deviations/SAEs |
Tech Stack
- Backend: Python 3.9+ / FastAPI / Uvicorn
- Frontend: Vue 3 / Pinia / Element Plus / ECharts
- Database: MySQL (production) / SQLite (dev) — auto-switch via env
- Vector DB: ChromaDB (persistent, semantic search)
- Auth: JWT (HS256) + role-based access (Admin / PV Specialist / CRA)
- Container: Multi-stage Docker build (Node → Python + Nginx)
Architecture
Three-layer separation:
Top comments (0)