I sit between the business question and the ML system — the part most teams get wrong.

A PhD in systems neuroscience taught me to anchor a nebulous problem into a testable hypothesis before writing a single line of code. 12 years of applied ML taught me to scale the answer once I have it: distributed training, large-scale retrieval, high-throughput inference. The result: fewer models built for the wrong reason, and faster paths to production for the right ones.

Currently at Databricksturning expertise into infrastructure — so any field engineer can deliver expert-level ML guidance, from distributed training to inference throughput.