Work
Four disciplines, one operating mode.
Most of this work comes from Hybridge, a dental implant company where I've been the only engineer in the building. Every project there runs the full arc: find the real problem, design the system, train the model or wire the LLM, ship it into someone's workflow, and teach people to run it. Different projects lean on different disciplines. Here's the work, told four ways.
AI Engineering
I build LLM pipelines for healthcare, where a made-up sentence isn't a quirk, it's a liability. Every system I ship forces the model into a schema, makes it cite its evidence, strips patient data before inference, and routes anything uncertain to a person.
-
-
-
Forward-Deployed Engineering
At Hybridge, a dental implant company, I've been the only engineer in the building. That makes every project forward-deployed: sit with the person who has the problem, dig the requirements out of their actual data, ship into the tools they already use, then train them until they don't need me in the loop.
-
-
-
Machine Learning Engineering
I train and ship models with the unglamorous parts done properly: versioned data pipelines with no train/serve skew, holdout gates that block bad models from promoting, experiment tracking, and honest evals. Flagship: a medical-imaging system that replaced a $98K vendor quote for about $25 a month.
-
-
-
Software Engineering
Every AI system I ship is wrapped in software built to last: databases that enforce their own rules, codebases held to strict typing and more than a thousand tests, Terraform for everything, and runbooks so the next engineer can pick it up without me.
-
-
-