Machine Learning Engineer, Turbine Performance

About this role

As a Machine Learning Engineer on the performance team you will own the core modeling problem: establishing what a specific turbine should have produced in the conditions it actually experienced, and deciding when the difference between that and reality is a fault rather than weather.

The constraint that shapes every design decision here is that a false positive has a physical cost. A finding sends a crew up a tower, and offshore it commissions a vessel. Precision is not an academic metric in this domain, and a model that flags everything is worth less than no model at all.

What you'll do

• Own the per-turbine expectation models end to end, from data and features through evaluation and production.
• Build the detection layer for developing component faults, and the evidence that accompanies every finding.
• Design evaluations that hold up on fleets with no labeled ground truth, which is most of them.
• Quantify each finding in annual energy and revenue, and be able to defend the derivation.
• Monitor models in production for drift as fleets age, components are replaced, and conditions change.
• Work directly with the data engineering team, whose output structure is your input contract.

What we're looking for

• 5+ years building and shipping machine learning systems in production, with real ownership of models other people acted on.
• Strong applied background in time-series modeling, anomaly detection, or forecasting.
• Expert in Python and the modern modeling stack, with the ability to work below the framework abstractions.
• The discipline to measure what you ship, including the cases where the honest answer is that the model cannot tell.
• Experience communicating a model's reasoning to people who will be accountable for acting on it.
• Comfortable owning open-ended production problems on a small team with broad technical scope.

Nice to have

• Experience with physical asset data: energy, manufacturing, aviation, or heavy industry.
• Background in reliability engineering or prognostics and health management.
• Familiarity with wind resource modeling, power curve analysis, or wake modeling.

This role is for you if you are

• Would rather report five findings that hold than fifty that need triage.
• Can move from an ambiguous question about a fleet to a measurable evaluation and reliable production behavior.
• Works effectively on a small team where technical judgment, direct collaboration, and production ownership matter.