About this course
Go beyond “chunk, embed, retrieve.” This course covers parsing messy documents, hybrid search, reranking, query rewriting, corrective loops, citations and — most importantly — how to measure whether any of it helped.
What you will be able to do
- Parse PDFs, slides and spreadsheets reliably
- Combine keyword and vector search with reranking
- Answer with citations and refuse when evidence is missing
- Build an evaluation set and track quality over time
- Deploy a RAG service with caching and guardrails
Curriculum
01RAG anatomy
- Where RAG fails
- Baseline pipeline
- First eval set
02Ingestion that works
- Parsing formats
- Chunking strategies
- Metadata
03Retrieval
- Dense vs. sparse
- Hybrid search
- Reranking
04Query side
- Rewriting
- HyDE
- Routing to SQL
05Self-correction
- Corrective RAG
- Self-RAG
- Refusals
06Evaluation
- Faithfulness
- Context precision/recall
- A/B testing
07Production
- Caching
- Guardrails
- Monitoring
Tools you will use
Prerequisites
- Python
- Basic LLM usage
Projects you will build
5 milestones
Airline customer support RAG chatbot
Answers baggage, refund and check-in questions from an airline’s published policies — with the source link for every answer.
LangChain · FAISS · Streamlit Coming soon
5 milestones
HR policy assistant with hybrid search on Google Cloud
Employees ask about leave, travel or benefits and get a short answer with the exact policy clause — across PDFs, Word files and slide decks.
Qdrant · BM25 · Cross-encoder reranker Coming soon
5 milestones
Clinical guideline Q&A with LlamaIndex
A grounded question-answering system over clinical guidelines and medical literature — from parsing to evaluation to a deployed API.
LlamaIndex · LlamaParse · Qdrant Coming soon
5 milestones
Enterprise advanced RAG: hybrid search, HyDE, self-correction and Text-to-SQL
Every advanced retrieval technique in one system, with an evaluation harness that shows which ones actually help on your data.
LangGraph · Elasticsearch · Qdrant Coming soon