Applications
RAG
Chunking, hybrid retrieval, reranking and how to measure a retrieval pipeline.
26 of 15 planned
L1 4L2 11L3 8L4 3
RolesLLM Product EngineeringAI Evaluation and ReliabilityAI Data SystemsCustomer AI DeploymentDeveloper Experience and Education
- Retrieval or fine-tuning for the policy wiki
- The first metric to read after a wrong answer
- The retriever model and the reranker model
- Why vector search ranks by cosine
- Count the work of a reranker over the whole index
- Faithfulness and answer relevance grade different things
- Filter on metadata ahead of the vector search
- Read the parent-document retrieval function
- Read the rank fusion function
- Review a chunking choice before the first index build
- Settle a chunk overlap argument with one ablation
- The reranker is perfect offline and worse in production
- Vector search misses the article with the exact error code
- Which query rewriter fits a two-topic question
- Why hybrid search fuses ranks and not scores
- Build the eval that picks a retrieval configuration
- Design a question answering service over company documents
- Faithfulness stays high while reviewers find invented facts
- Let the assistant say it does not know
- Pick how many candidates the reranker scores
- Read the four lines that score a BM25 term
- Recall holds, MRR drops, nobody touched the retriever
- The invented passage made code search worse
- Design code search across ten repositories that stays fresh
- Every metric is high and one unit still gets wrong answers
- The other jurisdiction's clause reached the answer
No item matches this level.
Taught in
The answers on this card come from these lessons.
- Question Answering Systems
- Chunking Strategies for RAG
- RAG (Retrieval-Augmented Generation)
- Advanced RAG (Chunking, Reranking, Hybrid Search)
- Capstone 02
- Capstone 08
- Chunking Strategies, Compared
- Hybrid Retrieval with BM25 and Dense Embeddings
- Cross-Encoder Reranker
- Query Rewriting: HyDE, Multi-Query, and Decomposition
- RAG Evaluation: Precision, Recall, MRR, nDCG, Faithfulness, Answer Relevance
- End-to-End RAG System