Vol II · Section 2
Chunking & Embeddings
The least glamorous section and the one that decides your ceiling. No reranker saves a bad chunk boundary.
1How would you choose the optimal chunk size?2What is the impact of chunk overlap on retrieval?3How would you chunk tables, PDFs, and structured documents?4What problems occur when a document is chunked without preserving context?5How would you handle documents containing images and text together?6How do different embedding models affect retrieval quality?7When would you use separate embeddings for queries and documents?8How would you detect that your embedding model is the bottleneck?9How would you migrate to a new embedding model in production?10How would you maintain embedding consistency when documents are updated?