I led the org that is responsible for serving and inference across Facebook, Instagram, and Threads. Our billions of users depend on us to deliver the most engaging results in their feed powered by advance AI models. We optimize and transform models from our ML engineers to efficiently inference across a heterogeneous fleet of AMD, Nvidia, and Meta accelerators in order to rapidly scale up our GPU fleet to meet the rising demand. We redesigned our serving system to dynamically scale query complexity based on real-time capacity headroom to capture excess compute during off-peak hours. Our investment in agentic AI enabled us to increase reliability while reducing production support load on our engineers.

I led the IGML org at Instagram. We are responsible for providing the end to end training and serving infrastructure for ML teams across Instagram, including feature authoring, training data construction, distributed training on GPUs, post training optimization, as well as the entire serving stack. We rapidly scaled up GPU adoption, growing our fleet by nearly 100k GPUs, increasing our models to hundreds of gigabytes, and drove ML strategy, such as large-scale foundation models, real-time streaming model updates, etc., resulting in large engagement wins.

I led a globally distributed SRE organization spanning across three continents. We were responsible for some of the world's biggest storage services, and a top revenue driver for growing the Google Cloud business. The Google Cloud Storage and Databases product area included object storage (Cloud Storage/Blobstore), block storage (Persistent Disks), file storage (Cloud Filestore), relational databases (Cloud SQL), and NoSQL databases (Cloud Memorystore). We focused on enhancing user trust, and ensuring that we provide the most scalable and reliable platform for our customers to grow their business.

I led all of Search Engineering at eBay. My team was responsible for the end-to-end Search experience, including the frontend UI, the ML algorithm that ranks which results to show, and the search engine that powers everything. We were consistently the biggest growth driver at eBay, and was responsible for more than 80% of eBay's revenue. We built our own Cassini search engine from scratch, in order to innovate new shopping experiences, such as image-based search, guarantee delivery time from third-party sellers, and personalized search results. We continue to aggressively grow our team, adding dozens of new engineers each year, while sustaining a culture of excellence.

I drove the development of hot social features on Bing, such as personalizing your search results based on what your Facebook friends have Liked. I co-architected a new indexing architecture that handle billions of personalized queries, retrieving Facebook pages liked by your friends. I also designed and implemented a Twitter recommendation engine, suggesting relevant Twitter accounts to follow based on your interests.