I'm a software engineer interested in systems and how they influence the shape of software.
I currently work on ML Infrastructure at Affirm and live in San Francisco.
My motivation is to build software that has a material impact on how people explore ideas.
Work
I joined Affirm with the impetus of working on ML and data problems at scale. My core contributions include:
Developing a new ML inference stack for transformers and other model types.
Leading early designs of Affirm's ML feature store, focusing on offline backfills over 100 TB+ data and latency forecasting for online flows.
Redesigning Affirm's experimentation platform, one of our largest surface area services, to support 1B+ daily evaluations while preserving core analytical models.
Prototyping AI assistants for customer support back in early 2023, later leading to the development of a new ML team.
Building streaming infra that powers risk calculations for every Affirm Card user.
And other fun stuff: writing interview questions, scaling teams, BFCM on-call shifts.
Prior to Affirm, I worked on systems at NASA JPL and research at Brown University.
Projects
Tahoe (2025). Trained a series of low-budget LLMs from scratch on OpenWebText data. I particularly enjoyed applying Chinchilla scaling laws to guestimate token budgets and optimizing BPE tokenization.
Seeing Numerics (2025). An exercise in visualizing various floating point formats, and aims to be like musical scales but for systems.