Drop Vikram into Claude and get a search relevance engineer who builds the judgment set before touching a single boost, because untested relevance tuning is just superstition. Vikram makes search return the right thing: query understanding across tokenization, stemming, spell correction, synonyms, query classification and intent detection; lexical retrieval and BM25 tuning in Elasticsearch, OpenSearch and Solr; vector and hybrid retrieval with reciprocal rank fusion; learning to rank including feature engineering, LambdaMART and neural rankers; relevance evaluation with NDCG, MRR and precision at k, judgment lists, human raters and click models; online metrics and interleaving experiments; position and presentation bias; personalization boundaries; zero-result and long-tail queries; faceting and filtering; and the honest tradeoff between index size, query latency and quality. What you get →Query understanding: synonyms, spell correction, intent →BM25 tuning plus vector and hybrid retrieval with RRF →Learning to rank and neural rankers in production →NDCG and MRR judgment sets, interleaving, click-bias correction 📄 vikram-search-relevance-engineer.skill Under 2 min install Works with Claude, ChatGPT & any AI chat How to install Download the .skill package → open Claude → paste SKILL.md into your Project Instructions or system prompt → describe your requirement → Vikram builds the answer. Includes a full worked example so you see exactly what you get.
