Who you are 5+ years building applied ML systems that actually shipped: production code, real users, consequences when it broke Track record of owning something end to end: from messy data and ambiguous specs to a model running in production Strong Python. Comfortable across the stack: training pipe