Apr 2024 – Aug 2025 · 1 yr 4 mos · Gurugram, Haryana, India · On-site

• ‣ Enhanced an AI-powered developer tool by integrating tree-sitter for real-time code analysis and leveraging AgentSpace and OpenSearch to unify organizational context across Confluence and GitHub.
• ‣ Unlocked compute and cost savings by applying autonomous Al agents with deep research to analyze large codebases, blending static analysis with runtime profiling to surface high-impact optimizations.
• ‣ Designed an AI Gateway enabling secure and scalable access to LLM providers, featuring:
• Dynamic load-balancing using real-time metrics to route traffic to the best performing endpoints.
• Fine-grained rate limiting and quota management.
• Schema standardization allowing zero-downtime model switching.
• ‣ Built semantic search service upon AWS OpenSearch:
• Implemented multi-vector indexing, hybrid search, custom scoring, and embeddings caching for performance at scale.
• Benchmarked index algorithms, sharding strategies, and search engines to guide architectural decisions - achieving 10k RPM at P95 latency of 400ms over a 100M+ documents index.
• ‣ Prototyped a natural language search over Zomato's catalogue enabling semantic search for complex long-tail user queries like "birthday party food, veg only, Jain."
• ‣ Enhanced restaurants search experience through distance filtering, veg-mode filter, and integrated cart planner for large group ordering.

Jul 2022 – Feb 2024 · 1 yr 7 mos · Bengaluru, Karnataka, India

• ‣ Engineered a high-performance ad-serving layer, achieving 2,500 QPS with P99 latency of 40ms, ensuring scalability and reliability under high traffic.
• ‣ Handled a 50x spike in traffic gracefully, avoiding service disruption and saving $100K/month, while enabling migration that saved $50K/year within 2 weeks.
• ‣ Developed a retailer management platform (internal tool), streamlining collaboration with advertisers, management of ad placements, and ad campaign flow controls.
• ‣ Overhauled campaign and asset management services, eliminating third-party vendor and adhering to clean code principles, reducing maintenance overhead.
• ‣ Optimized video loading performance by restructuring media files and reducing load times, while also introducing capability of dynamic post loading to improve user experience.
• ‣ Improved system observability with dynamic metrics collection, alerts, and traces logging, making issues easier to detect and resolve.
• ‣ Reduced infrastructure costs by 10x by switching from Redis to Aerospike for caching in high-load services.

May 2021 – Jul 2021 · 2 mos · Bangalore Urban, Karnataka, India · Remote

• ‣ Developed a scalable OTP automation system using GSM modems and SMS listeners, enabling seamless access to secured web services. Boosted workflow efficiency with regex-based entity extraction and AWS data pipelines.