Jan 2024 – Present · 2 yrs 2 mos · Bengaluru, Karnataka, India · Hybrid

• Integrated server-sent events (SSE) in scalable chat APIs to handle high traffic, ensuring low latency, and
• increasing engagement by 10%
• Optimized multiple Chat APIs by integrating Python generators with SSE, enabling real-time streaming responses that significantly reduced latency and boosted user retention
• Integrated Google Pub/Sub to stream agentic AI thoughts to the frontend in real-time, enhancing user experience and enabling seamless visualization of multi-agent reasoning processes.

Jul 2022 – Dec 2023 · 1 yr 5 mos · Bengaluru, Karnataka, India · Hybrid

• Spearheaded the optimization of table summarization system using LLMs,
• leveraging asynchronous operations to achieve a 70% reduction in response
• time, significantly improving user experience.
• Developed recommendation APIs to get relevant questions based on
• user-selected tables utilizing concurrent operations, caching, and other
• optimizations leading to improved API performance by 80%, ensuring an
• efficient user interaction.
• Implemented a micro-services architecture to enhance the performance of
• a recommendation API by offloading processing tasks, including Redis task
• scheduling and caching from main API resulting in a significant reduction of
• over 65% in response time
• Increased vector search precision by 58% based on user-selected criteria
• within the vector store, optimizing retrieval for RAG search and therefore
• improving the relevance of search results.
• Implemented concurrent processing for expensive database operations in
• python to solve scalability challenges.
• Devised algorithms to track pipeline states, including success and failure
• due to various factors, and provided real-time status updates to BigQuery
• housekeeping tables minimizing downtime by 19%.
• Conceptualized corrupt file routing algorithms, enhancing pipeline
• robustness by segregating inconsistent/junk files into a separate bucket
• leading to optimized resource utilization and reduced processing
• overheads by 14%.