I'm a software engineer and researcher currently based in San Francisco, CA. My interests lie at the intersection of machine learning and systems. I'm currently a Member of Technical Staff at Databricks, where I work on systems for large language models.

I studied computer science at Stanford, focusing on systems and artificial intelligence. I wrote my honors thesis on using layer-based editing interfaces from graphics to improve control of image generation models, advised by Prof. Kayvon Fatahalian and Prof. Christopher Ré. My recent projects include:

• Collage Diffusion

  A fully open-source layer-based editing tool built off Stable Diffusion turning composited images into harmonized, consistent collages.

• Multi-Object Multi-Actor (MOMA)

  A dataset containing 148 hours of annotated videos, MOMA presents a compositional and hierarchical structure to analyzing complex interactions in video.

Previously, I worked with Prof. Fei-Fei Li and Prof. Ehsan Adeli at the Stanford Vision and Learning Lab on a large video dataset of human activities. I've also worked on infrastructure to quickly and cheaply train billion parameter large language models at MosaicML and training large models with the Megatron library at NVIDIA's Applied Deep Research Team. Long before that, I was part of the team winning the World Schools Debating Championship in 2018.