Abhinav Tyagi

• Led architecture and backend ownership of the customer-facing Search Platform handling 25K+ RPM, orchestrating low-latency Elasticsearch querying, real-time stock validation, pricing computation, marketplace-retail aggregation, voucher applicability, and geo/store-aware product availability. • Built an LLM-powered analytics agent to identify and generate reports for low-performing search queries by analyzing topproducts and query behavior, relevance improvements to enhance CTR and Add-to-Cart (ATC) rates. • Designed and developed distributed ingestion pipelines for Stock, Pricing, Marketplace Offers, Promotions, and Elasticsearch Sync using Apache Kafka, Azure Event Hub, MongoDB, Redis, and Elasticsearch, enabling near-real-time indexing and commerce data propagation for a catalog of nearly 10M products. • Developed intelligent semantic product-tagging pipelines and an LLM-based tagging system to improve search relevance, query-aware boosting, and intent-based product discovery for queries such as dairy-free, lactose-free, and vegan. • Built a Learning-to-Rank (LTR) framework using XGBoost/XGBRanker with lexical, semantic, category, brand, popularity, CTR, ATCR, and order-rate signals integrated into Elasticsearch rescore pipelines. • Led implementation of hybrid lexical + Two-Tower Vector Search and analytics-driven query tuning for broad, ambiguous, and long-tail queries. • Integrated NER-based query understanding to detect product quantities, sizes, and units directly from queries, improving filtering, ranking, and recommendations. • Designed scalable automated filter-integration frameworks and a Dynamic Taxonomy & Filters Platform with runtime-configurable PLP/SLP filters and admin tooling for category-aware merchandising without redeployments. • Built cross-CTA search functionality to dynamically surface faster-delivery inventory options and improve conversion through delivery-aware ranking strategies.