I'm a software engineer and systems researcher. My work spans two ends of the systems spectrum: pure distributed software, and the boundary where software meets hardware.
On the software side, I work on Pathways and Plaque — Google's distributed dataflow system for ML and the sharded runtime it's built on. Before that, I was the owner of a custom streaming data-processing system that operated at Google cloud scale while guaranteeing strong consistency at low latency.
At the hardware boundary, I worked on INSIDER , an in-storage computing platform, and ST-Accel for FPGA streaming, and earlier worked on ClickNP at Microsoft Research — part of the Catapult effort to bring FPGAs into the datacenter.
These days, I'm focused on scaling LLM pre-training systems — and just as curious about the payoff: turning raw capability into applications that matter. This is where I write about what I'm building and thinking about.