I'm an AI Engineer specializing in Healthcare IT, building production-grade systems that bridge hospital operations with modern AI capabilities.
My flagship project β HISDashboard β connects directly to a hospital's core HIS (Hospital Information System) via Oracle Database, providing AI-powered analytics, automated reporting, and multi-agent workflows that help hospital staff make data-driven decisions daily.
class DucThanh:
role = "AI Engineer"
domain = "Healthcare IT"
focus = ["Multi-LLM Agents", "RAG Systems", "Production AI", "Full-Stack"]
current_project = "HISDashboard β Hospital Management System with AI"
stats = {
"api_endpoints": 292,
"ai_agents": 10,
"llm_providers": 4,
"docker_services": 6,
}HISDashboard β Hospital Management System with AI
Full-stack hospital dashboard integrating AI agents, automated reporting, and real-time analytics β connected to a live Hospital Information System (Oracle HIS).
|
π’ By the Numbers
|
β‘ Key Technical Decisions
|
Architecture Overview:
User β React/Ant Design β FastAPI (292 endpoints) β MySQL + Oracle HIS
β
AI Agent v2 (10 ReAct Agents)
βββ RAG (ChromaDB + Hybrid Search)
βββ Multi-LLM (4 providers + fallback)
βββ MCP Protocol (Tool Server)
βββ 40+ Tools (Analytics, Finance, HR, Predictive)
βββ Bot (Telegram + Zalo auto-reports)
- π¬ AI Agent Orchestration β Multi-agent systems with reflection and self-correction
- π MCP Protocol β Building standardized tool servers for AI applications
- π Healthcare Analytics β Anomaly detection and predictive analysis for hospital operations
- π€ Automated Reporting β AI-generated insights delivered via Telegram/Zalo bots
"Ship production systems, not prototypes."
- π― Systems > Models β A reliable system beats a slightly more accurate model
- π Multi-LLM > Single-LLM β Never depend on one provider in production
- π Documentation is code β If it's not documented, it doesn't exist
- π₯ Domain expertise matters β AI without context is just a party trick
