New York, New York, United States
I'm the Overall Tech Lead & Hands On Lead Engineer for Meta AI on Facebook App - this includes Feed Deep Dive, Write with AI, AI-driven Search and several other AI features on Facebook - we power AI-driven experiences and serve the largest consumer use-case in the world. Personal code contributions focus on solving challenges with model quality, capacity and latency, i18n, multimodality and social graph understanding. Also, authored Automated Eval systems to speed up our product development cycles from multiple months down to a few days - with proxies for online and offline metrics for quality, engagement, trust and safety - keeping our users safe while delivering a highly engaging AI first experience. Our product drives a super-majority of usage for Meta AI in the US and in several other key markets.
I also lead the AI for productivity for Facebook - code authoring, testing, experimentation, design-to-code adoption and tool-building using agentic AI assistants. Personal code contributions involve App-scale automated test authoring using AI agent tools - also refactoring several core code-bases powering Feed, Stories & Entities.
Previously, I have tech led Facebook Feed Experience & Delivery (owning the Facebook Feed surface with 3.5 billion MAU). I have worked on capacity footprint reduction for Meta's fleet focussing on automating detection, measurement, fixing and improving the CPU & GPU capacity bottlenecks in Meta's low-level libraries to make data-center scale savings. Prior to that, as the Senior Engineering Manager at Instagram, supported a team of 45 staff and principal engineers focussed on optimizing app start latency, scroll performance, reduced crashes and app freezes driving some of the largest engagement metric improvement for our 2.5 billion user base.
(https://about.fb.com/news/2024/04/meta-ai-assistant-built-with-llama-3/)
(https://cppcon2023.sched.com/event/1REuc/open-content-jubin-chheda) https://arxiv.org/abs/2402.09171