about_

I’m super interested in inference engineering. I like figuring out how to make models run well when the hardware is limited, the network is unpredictable and someone has to operate the thing once it is deployed.

At Meta, I worked on the disaster recovery drain tests. I built an agent that automated planning six weeks of upcoming simulations, including which regions to drain. I also built a system that shared high-risk services with the team before each test.

I then built an eval framework around those predictions. It compared them with the SEVs that actually came out of a drain, helped separate prediction misses from service onboarding gaps and gave the team a dashboard to look through past tests and individual incidents.

Before Meta, I worked on AWS deployment infrastructure and CI/CD, built an LLM pipeline for inventory automation and contributed Kotlin and Android functionality to Pocket Paint through Google Summer of Code.

Outside work, I maintain Dictate, a native Android text-to-speech app with 90K+ installs.

Right now I am building EveryGPU, a distributed inference experiment across remote GPUs.