cs at uoft. building things, running experiments, occasionally writing about them.
website ↗ writing ↗ linkedin ↗
Working across AI infrastructure, model evaluation, and efficient inference. Some of that lives here. Some is still taking shape.
Determina
Testing what happens when you change an AI system. Recommenders, agents, simulated users, repeatable failures.
KV-cache research
Precision, memory, and long-context inference. Smaller doesn't always mean faster. Some experiments that didn't work out.
MedRAX-2 · GPCR-Claw
Medical imaging agents with WangLab. Protein-design workflows with GPU models. Different problems, plenty of infrastructure.
| spareNode | Put the spare Linux machine to work. Docker and GPU jobs over SSH. Early alpha. |
| anySDK-MCP | Turn Python SDKs into tools agents can call. |
| codex-commute | Close the laptop. Let the job keep running for a bit. |
Built endpoint and AI-security infrastructure at Forge. Worked on recommendation systems and the PS5 Store at PlayStation. Chess models and computational social science at UofT.
Co-founded GenAI Genesis. Contributed to Google ADK. Have taught some CS and math along the way.