Senior Software Engineer - Victoria's Secret & Co - Bengaluru, KA, India
(2025-03)
- Redesigned Redis read paths by replacing request-time fan-out with a Kafka-driven precompute pipeline partitioned by shop_id, introducing bounded worker pools, atomic pointer reads, and batched MGET to eliminate cache stampedes, reducing Redis load and stabilising low-latency reads under peak traffic
- Architected VogueFlow, a real-time multi-agent AI commerce engine for Victoria's Secret using Go, LangChain, Claude API, Pinecone hybrid RAG, and Redis — delivered less than 200ms p95 latency at 15K RPS via tiered semantic/token caching and model cascading (Llama-3.1 to Claude-3.5), cutting inference costs 42% (~$8K/day) and driving a projected 22% mobile conversion lift with $34 higher AOV through agentic product discovery and styling workflows
- Built an AI-powered observability layer on Prometheus using Prometheus MCP to translate natural language queries into PromQL. Enabled faster metric retrieval, service-health scoring, and AI-assisted incident analysis, reducing diagnosis time and on-call cognitive load
Software Engineer - Victoria's Secret & Co - Bengaluru, KA, India
(2023-04 - 2025-03)
- Developed an optimised Add to Bag flow from the Product Listing Page for Beauty and Sleepwear collections. Increased annualised sales by $2.3M by reducing purchase friction and improving conversion
- Led migration of 120+ microservices from Consul to Harness and ArgoCD GitOps pipelines. Coordinated a BLR cross-functional team to deliver 425 service versions across 2,500 PRs, modernising CI/CD and removing manual deployment errors
- Designed and built end-to-end observability dashboards in Prometheus and Grafana, cutting on-call troubleshooting time by ~50% and improving regression detection
- Optimised peak-load performance by refactoring database access patterns and removing redundant Kafka retries, reducing P95 response times by 30%
Software Development Engineer - Biofourmis - Bengaluru, KA, India
(2020-05 - 2023-03)
- Architected event-driven ingestion pipelines in Golang, Kafka, and AWS Lambda to process 400B+ daily biometric data points across 100K+ patients, improving ingestion throughput by 5× and reducing clinical alert latency by 40%
- Built scalable microservices to collect and process continuous biometric data streams, improving reliability and supporting high-volume clinical data processing
- Designed event-driven pipelines using Golang, Kafka, and AWS Lambda, integrating PostgreSQL and MongoDB to support structured metrics and semi-structured vitals at scale
- Improved ingestion throughput by 5x and reduced clinical alert latency by 40% through pipeline redesign and data-store optimisation
- Engineered a configurable care-execution workflow engine supporting 50+ clinical protocols and maintaining sub-250ms p95 response times under peak load for 100K+ concurrent patient sessions
