Converting code from one ML framework (like PyTorch) to another (like JAX) while preserving functionality.