Design, optimize, and implement features that leverage LLM to automate complex tasks.

Implement automated evaluation frameworks to assess the quality of LLM-generated output.

Research, experiment with, and integrate the latest cutting edge AI techniques.

Conduct quality audits of LLM features and drive product improvements with comprehensive A/B testing

Implemented proximity search in Command to aggregate documents that fall within a geographic region using Redis caching, Mapbox geocoding, and Geopy computations

Added language-awareness to Primer Analyze to support the Chinese and Spanish experiences

Integrated Asyncio into the ingest pipeline to improve translation efficiency in the create-and-poll mechanism

Wrote C++ APIs that utilize multi-threading and relation predicates for surfacing resource status information of related entities in one place for handling appeals

Improved reliability of Python verification logic by restructuring the code to separate sync and async flows

Configured data pipelines and SQL scripts which run on large datasets to create aggregate signals that provide safeguards when taking policy-based actions

Conducted live user A/B testing experiments with prospective features and performed deep dive data analysis using Google Colaboratory (SQL, Python) to investigate various engagement metrics and identify launch candidates

Launched C++ Mix filters with extensive end-to-end testing that check policy restrictions to ensure content safety for young at-risk users

Designed and implemented C++ client-side components with comprehensive unit, integration, and functional tests that utilize RPC mocks to support migrations that affect live production workflows

Wrote APIs that perform reads and writes against a version-controlled data management system

Leveraged JUnit and Mockito to write extensive unit, integration, and functional tests