Work
A few scenarios
Two of these are real client work (anonymized). The AI Operationalization scenario is still illustrative, and will be replaced with a real case study as that kind of project wraps — marked clearly until then.
Client work — anonymized
Strategy Workshop
ProblemA national extended warranty company had a long list of possible AI/ML use cases and no way to know which were actually worth pursuing.
What I didRan a single strategy day — worked through the full list, narrowed it to the two with the most promise, and scoped both so the team understood the cost and effort to get each to production.
ResultBy the end of the session, the team knew what data needed to be in their warehouse, what “clean” meant for their use cases, which platform to build on, and enough MLOps fundamentals — taught at a high level — that everyone in the room understood the process ahead.
Client work — anonymized
ML Operationalization
ProblemA flavor and fragrance manufacturer had a chemical analysis model that worked in a notebook but had never made it to production.
What I didTook the model from notebook to full production deployment on Snowflake, training their chief data scientist on the platform as the project progressed.
ResultModel training time dropped from 48 minutes to 22 minutes — more than half — with no drop in model metrics. Their chief data scientist came out of the project with a real working understanding of MLOps on Snowflake for whatever comes next.
Sample scenario
AI Operationalization
ProblemAn internal copilot that answered confidently — including when it was wrong, which security wouldn’t sign off on.
What I didAdded a verified-query semantic layer, guardrails against unsafe outputs, and an automated eval pipeline.
ResultSecurity signed off, and the copilot is now in front of real users instead of stuck in pilot purgatory.