Architecture review & code audit
A close read of an existing system: the risks, what's worth changing, and the smallest changes with the biggest effect.
12+ years building production systems. Independent research on efficient inference and quantization.
Fixed-scope work, done by me end to end. You get written deliverables you can keep, not hours on a timesheet.
A close read of an existing system: the risks, what's worth changing, and the smallest changes with the biggest effect.
Running open models on your own hardware: sizing, quantization, serving, and an evaluation on your own tasks before you commit.
Using LLMs to make decisions rather than write text: routing, classification, scoring and guardrails, with a test set to measure them.
A few hours a month of async reviews and second opinions, for teams that want an outside senior engineer to check their thinking.