I built an open-source system that uses large-scale evolutionary search to discover governing differential equations directly from time-series data. It runs fully on GPU via PyTorch, supports deterministic checkpointing and resume, and is designed for symbolic physics and dynamical system identification. Would love feedback from anyone working in scientific ML or symbolic regression.