I work on user authorization products (API + webapp) for end users and Plaid's data partners / customers. Check out my.plaid.com to get a sense on how Plaid pioneers user authorization in the Open Banking space.

Technical leader in developing the control plane for Lyft’s Data Platform:• Provides managed analytical query executions in Hive, Trino, and Spark at a scale of 250K queries per day and an average of ~1K concurrent executions• Supports core data platform features like rate limiting, traffic shadowing, query history, and search• Lyft's Data Platform powers everything from pricing/ETA to mapping to GDPR compliance1 of 3 engineering leads in Hive Deprecation Task Force:• Responsible for minimizing manual effort in migrating 1K Hive ETL workflows (20K+ tasks and ~100K distinct queries) by building automated migration tools for Trino and Spark• Deprecating Hive and migrating to Spark/Trino increases ETL efficiency (faster query execution with higher reliability) and saves $3MM/yr in infra costs

• Enabled data-driven product decisions by creating pipelines to ingest, transform, and aggregate PB-scale datasets• Improved data reliability by building quality control tooling• Stack: PySpark/Hive on HDFS/S3/Greenplum, Python, and Oozie• Conducted 80+ technical interviews and organized on-campus recruiting events

• Created 1.5MM/year revenue impact by building pricing pipelines for new categories of structured securities• Provided more reliable data snapshots to 1.5K client institutions by improving existing pricing services for 1.3MM structured products• Stack: C++ and Python

Selected as 1 of 34 participants out of 400+ applicants for the inaugural GGV Fellows program, which was designed as a close-door intensive training on entrepreneurship and technology in China. The program was taught by a group of 20+ founders and top executives from some of the most valuable technology companies in China, including ByteDance, DiDi, Xiaomi, Xiaohongshu, Megvii, and Momenta.

CS Theory, AI, Intro to Databases

• Improved causal relation detection rate on a Wikipedia corpus• Improved parsing for Abstract Meaning Representation