Steven Milanese

Calibrated decision models over closed answer sets · Rust-native inference and training (Burn) · evaluation harnesses and honest scorecards · calibration, order-invariance and injection robustness · agent orchestration with human gates · local-first, on-device ML