MediumData

Implement a simple retriever

1Give yourself 5 minutes

2Answer out loud, not in your head

3Then compare with the answer below

Reference answer

Naive baseline: TF-IDF with cosine over tokenized docs, returning top-k. Vectorize: encode docs with sentence-transformers/MiniLM, store in a FAISS flat index, query via dot-product, return top-k by cosine. Senior signal: discuss pivot from flat to HNSW at a million vectors; ground the design with semantic chunking (200-500 tokens, 50-token overlap) so chunks stay coherent. Discuss query rewrite/expansion, hard negatives mining. Discuss evaluation: recall@10, nDCG@10 on a held-out query set.

Then expect these follow-ups

How would you cache this?
Cross-encoder reranking; what about ingest cost?

Free to read · better with Enzo

Practice this out loud with Enzo

Enzo runs it as a mock interview, pushes back with follow-ups, and grades you on the rubric.

Continue with Enzo Sign up free

Next question

Implement a simple retriever · Interview Question | landed