Jun 2022 – Present · 3 yrs 11 mos · Bengaluru, Karnataka, India

• Lead backend engineer working on large-scale discovery and ingestion platforms that power high-traffic, revenue-critical user experiences at Meesho.
• Led the org-wide re-architecture of the MARVEL ingestion and ranking pipeline, setting long-term technical direction by migrating the system from an API-driven model to an event-driven architecture using Apache Flink 2.0 and Apache Paimon. The redesign significantly improved resilience, decoupling, and failure isolation, and enabled change-data–driven processing across downstream consumers.
• Drove cross-org technical alignment across 7–8 teams, planning execution milestones, delegating components to junior engineers, and guiding implementation through design reviews and architecture forums.
• Led large-scale performance optimizations across Organic Feed Service (OFS) and Sort & Filter systems, dramatically reducing tail latency and improving overall query efficiency through Elasticsearch tuning, Redis optimizations, and large-scale data cleanup.
• Reduced Elasticsearch storage footprint substantially by pruning invalid catalogs and cleaning up index mappings, unblocking long-stalled indexing pipelines and improving system operability.
• Own and support multiple business-critical discovery and commerce systems including Instant Checkout, Meesho Balance, Hero PID, Affiliate indexing, Parallel Feed V2, and More-Like-This, ensuring correctness and reliability during peak traffic and sale events.
• Enabled Meesho’s first regional sale capability by introducing cohort-based ranking in CLP/OFS, achieving high cache efficiency and preventing incorrect nationwide boosting during regional sales.
• Scaled Gold Feed-in-Feed (FIF) to handle high peak traffic by clearly separating embedding-based and Elasticsearch-based pipelines and defining robust scale policies.