Gemini for Google Workspace core platform.Gmail & Chat machine learning infrastructure.

Research engineer on the Speech team.Worked on developing sequence-to-sequence deep learning models for speech recognition in long-form conversational settings.

- Full-stack developer for CodaLab, an open-source platform for machine learning researchers to organize and collaborate on experiments.- Redesigned and reimplemented the backend infrastructure from the ground up, replacing a monolithic XML-RPC-based system with a transactional REST API backed by a microservices architecture, building out an OAuth 2.0 authentication flow, and migrating all frontend clients with zero downtime. - Stabilized codebase with various features and bug fixes, closing 120+ public GitHub issues.- Ran a competition on a SQuAD reading comprehension task using CodaLab for the CS224N NLP course with 162 student teams.

Created smart home events stream for training machine learning algorithms (e.g. occupancy detection) by integrating events pipeline with Apache Kafka.

Built an extensible chatbot that allows DevOps engineers to trigger certain actions, such as the deployment of AWS EC2 instances, from Slack chat rooms.

Developed and prototyped multi-armed bandit system to optimize budget allocation for real-time bidding across different ad creatives in an ad campaign.

Led a weekly discussion section, held office hours, and graded for the introductory CS course. Topics included basic data structures, software engineering, and object-oriented programming.

Built web application for a "cloud" biology platform that allows users to interact with live organisms remotely in real-time. Used subsequently in a bioengineering lab course at Stanford.