LLM Engineering for Healthcare
From prompt engineering to production-grade clinical AI systems
A graduate-level course covering the full stack of building LLM-powered systems in clinical and biomedical settings - from foundational transformer architecture to RAG pipelines, agent design, hallucination evaluation, and HIPAA-aware deployment on Azure.
Attention mechanisms, tokenisation, context windows. Why GPT-4 behaves differently from Llama 3.
Zero-shot, few-shot, chain-of-thought. Structured output with Pydantic. Clinical prompt patterns.
Deploying GPT-4o on Azure. Token budgeting, rate limiting, logging for compliance.
FAISS, Chroma, Pinecone. Embedding models for biomedical text. Chunking strategies.
HyDE, re-ranking, self-RAG, multi-hop retrieval. RAGAS evaluation framework.
Processing EHR notes, discharge summaries, clinical guidelines with LangChain + Azure.
ReAct, tool-use, OpenAI Assistants API. Building clinical decision support agents.
Faithfulness, answer relevance, context precision. Building evaluation pipelines with RAGAS and custom benchmarks.
Data residency, PHI redaction, audit logging, model version control for clinical AI.
Containerisation with Docker, CI/CD for ML, monitoring with Application Insights.
Teams build a full RAG + agent pipeline for a real clinical use case. Peer-reviewed presentation.