Abhishek Shirgaokar Email & Phone Number

Bengaluru, Karnataka, India

- Developed real-time personalized recommendation service for the ”ForYou” section of a short video platform with 2M+ DAU, achieving sub-300ms p95 latency and driving a 150% increase in the north star metric.-Implemented self-managed PostgreSQL cluster on EC2 instances with Instance Store, replacing the previous video watch history database setup. Achieved over 90% reduction in cloud costs while maintaining performance through hardware selection, data partitioning, and query enhancements, leading to Index Only scans on the database.

Bengaluru Area, India

Charmboard is an AI startup leading the shift from search to discovery online by enabling visual discovery of fashion from popular culture, primarily videos. My major work at Charmboard involves revamping the flow of detecting and identifying all unique fashion looks from a video. Some of my awesome projects -> Applied density based clustering in multiple embedding spaces, each embedding generated from a custom feature extraction network, to greatly improve cluster quality and reduce manual QA in the process unique look identification and enabling touch-to-discover feature in a video. Leveraged multiple server-less container execution (GPU & CPU) solutions and their orchestration on public cloud to build horizontal scaling capability as well as portability in the solution.