Hi! This is the wonderfully designed website of Johnston Liu. I'm from Vancouver, Canada, and I'm a machine learning engineer at Suno based in Cambridge, MA. I stare & listen to data, read claude, and make gpus do things. I'm a graduate of the University of Chicago with a double major in Mathematics and Computer Science.

Prior to moving to machine learning, I worked on our infrastructure team. I spent my time on:

  1. On-call incident triage and Slack engineering agents (credentialless sandbox w/ bash + cli tools, integrated with existing company tooling, harness-outside-sandbox runtime and Durable Object-style SQLite WAL / S3 state).
  2. Kubernetes sandbox platform (gVisor isolation, Cilium and Envoy egress controls + credential injection, ~1s cold starts via node image prepulling, warm node buffers, and a node-local CSI git cache).
  3. Persistent Cloud Hypervisor microVM devboxes (Tailscale, checkpoint/restore migration, live-swappable VM runner upgrades, memory hot-plug, and live resizing).

Before all of this, I worked in the Bay Area at Pure Storage on the FlashBlade kernel team getting kernel modules and hardware devices to work. I also built an AI Agent (maybe at some point i'll write this, but quickly became obsolete). Before that, I was in NYC at CMS Holdings performing research into trading strategies and building research tools in Python. And even before that, I did mathematics research on symmetric monoidal preorders and their applications to finite topological spaces. The relevant paper (archaic at this point) can be found here. If you'd like to contact me, feel free to use any of the links below!