About
I work on search — specifically the part between someone typing a query and getting back something they actually wanted.
This blog is where I think through that problem in public: hybrid retrieval, query understanding, reranking, and where search and AI are converging. Mostly posts I wish I’d found while figuring something out.
On Walmart’s search runtime team I’ve worked on the recall and ranking path of ecommerce search at scale — re-architecting the L1 recall stage to merge lexical candidates from Solr with an ANN dense index into a single set for a cross-encoder reranker, query fan-out and result-merge in the search orchestrator, and a lot of unglamorous Solr tuning, which is where much of the latency budget is won or lost. Before that, at Tracxn, I ran large-scale indexing and query serving on Elasticsearch.
What pulls me toward relevance as a specialty is that it’s equal parts art and science. The science is measurable — judgments, offline metrics, telling a real improvement from noise. The art is everything the metrics miss: what a shopper actually meant, and how that shifts from person to person once personalization enters.
Right now I’m going deeper on embeddings, vector databases, hybrid retrieval and RAG — currently building a research-dossier system over Indian listed-company annual reports.
I’m in Bengaluru, and open to Lead or Senior engineering roles in search, relevance, or backend.