• Led a team of 4 engineers, transitioning the company’s core business logic from Postgres Stored Procedures to modern, typed Python (SQLAlchemy, Pydantic, etc)

• Partnered with SWEs, Support, and PMs to devise loan formulas and address discrepancies

• Created coding guidelines and drove use of modern tools like Graphite throughout Engineering

Software Engineer – PyTorch Compiler (github.com/pytorch/pytorch)

• Collaborated with internal, Google, and OSS engineers in building out Symbolic Shape Analysis for PyTorch / XLA integration.

• Built and maintained optimization passes for TorchScript, speeding up BERT by 25%

• Built a new system for delegating and writing release notes for PyTorch, halving the time it takes for people to prepare the notes.

Software Engineer – Integrity Measurement Platform

• Developed system to A/B test the impact new features on Facebook have on harmful content viewership using ML signals.

• Worked with 13 different client teams: building, maintaining, and debugging 50+ metrics.

• Mentored the onboarding of 4 Engineers and managed 2 interns for their internships.

• Taught two courses on neural networks: Convolutional Neural Networks for Visual Recognition (CS231n), and Natural Language Processing with Neural Networks (CS224n)

• Also taught Programming Languages (CS242), where I taught Rust, OCaml, and Lua

• Mentored hundreds of students through intro CS courses, teaching them how to think like a computer scientist in CS 107

• Built neural network for predicting credit card preferences (AUC 0.796 vs prod model’s 0.765)

• Developed web push notifications support for the Credit Karma website