My work

Most of my industry work has been in the broad region of research software engineering: implementing cutting-edge research into usable, tested, and correct software. I also enjoy more generally building the infrastructure and tooling surrounding and supporting new research ideas.

My resume is linked here, and more info is below.

March 2024 - Present: Software Engineering at Codeium.

I lead the training/eval infra team.

March 2023 - March 2024: Simulation Engineering at ReSim

I simulated robots in open-source C++, did scalable systems engineering in Go, and wrote metrics frameworks in Python.

February 2022 - March 2023: Software Engineering at Aurora.

I worked on synthetic world and sensor simulation for self-driving trucks, as one of three people on the core physically-based raytracing and renderer team, working with highly parallelized C++ code.

2020-2021: Graduate teaching assistant for Stanford’s CS 229: Machine Learning, and CS154, 254, and 254B: the full Complexity Theory track.

Summer 2021: Machine Learning Infra at X (formerly Google[x].)

I supported the Perception team at Mineral: X’s computational agriculture moonshot, working mostly with Tensorflow.

2019-2021, including Summer 2020: Computer science research in Chelsea Finn’s IRIS lab.

I worked on meta-learning module homomorphisms. So given a bunch of data from different functions which are all homomorphisms of the same module, how can you disentangle the “module symmetry” from the “task-specific parameters”? We had a paper on the topic accepted to ICLR 2021: Meta-learning Symmetries by Reparameterization.

I also worked on transferring invariances across classes, and improved optimization methods for multi-task RL algorithms.

Summer 2019: Machine Learning at Intuitive: the leading company in robotic surgery.

I was on the Advanced Data Analytics team, working on implementing machine learning research in NLP and weak supervision, working mostly with PyTorch.

Summer 2018: Software Engineering at Hulgrave: an early FinTech startup by the team behind Nutmeg. (Sadly, Hulgrave is no longer around.)

I built Hulgrave a small set of production-level algorithms for portfolio optimization, and machine learning algorithms for outlier detection, working with a mixture of Java and Python. I successfully pitched this product to a major investment manager.

Spring 2017: Research Software Engineering at The Alan Turing Institute, the British National Institute for AI and Data Science.

Highlights included working on neonatal infant health, contributing to the Alexa SDK, and co-authoring a talk at the UK Conference of Research Software Engineers, entitled Curating Datasets for Development of Automated Data Wrangling Tools.

Earlier: Work experience in biotech and at a hedge fund! Thanks to folks who gave me those early opportunities to get in the way.