Build a Document Search Engine

• ✓Build an inverted index and answer keyword queries over thousands of documents
• ✓Rank results with TF-IDF and BM25 and explain what each term in the score does
• ✓Measure a ranker with recall at k and mean reciprocal rank, with the spread across query samples, instead of eyeballing results
• ✓Test whether lowercasing, stop words, stemming, k1 and b help, and show how the choice of test queries changes the verdict
• ✓Handle typos with character n-grams and measure what that costs in memory and speed
• ✓Write a search note that states what the test cannot tell you

Build the retrieval layer under a search box or a "chat with your documents" system — an index, a ranker, and a test harness that tells you whether a change helped.