Building generative AI products for WhatsApp's user-facing chat experiences powered by LLMs

Worked in the Data Access team, designing and developing systems to handle the next phase of growth of TikTok's business and ensuring high stability and performance.

Returned to the Amazon Aurora team to help build the next generation databases for the cloud.

Worked on Aurora, a MySQL and PostgreSQL-compatible relational database built for the cloud.

Processing embedding layers involves the most memory-intensive steps in the inference of recommender systems that are used by platforms such as Facebook to make suggestions to their users. In this project, I am using Data Processing Units to increase the throughput of embedding operations in the inference of a recommender system.