Learn RAG and LangChain — without the demo-day shortcuts.

A course built around the RAG system you actually want to ship: your documents, your retrieval problem, your acceptance bar. Strive shapes the curriculum, streams lessons live, and the recall queue keeps the chunking, embedding, and evaluation choices straight when you revisit them six weeks later.

What you’ll learn

• Build a working RAG pipeline end-to-end — ingest, chunk, embed, retrieve, generate.
• Pick chunking and embedding strategies that match your documents, not a tutorial’s.
• Compare vector stores (pgvector, Pinecone, Weaviate, Qdrant) on the trade-offs that bite.
• Use hybrid search and rerankers to recover the recall a naive top-k loses.
• Evaluate retrieval and generation separately so you know which half is broken.
• Recognise when RAG is the wrong answer — long-context, fine-tuning, or just better prompts.
• Decide between LangChain, LlamaIndex, and a hand-rolled stack for your project.

A typical Strive course on RAG & LangChain

1. Why naive RAG demos lie — and what real systems look like3 lessons
2. Ingestion and chunking — choices that haunt you later4 lessons
3. Embeddings — picking a model and an index that match4 lessons
4. Vector stores compared — pgvector, Pinecone, Weaviate, Qdrant4 lessons
5. Hybrid search, rerankers, and the recall you’re losing4 lessons
6. LangChain and LlamaIndex — what they’re for and when to skip them3 lessons
7. Evals — retrieval metrics, generation metrics, and what users notice4 lessons

Demonstration outline — your course is generated around your answers, so module count, depth, and difficulty will differ from this. Across the 7 modules above there are 26 lessons.

Frequently asked