When I first started working in scientific machine learning, I understood the physics much better than the coding. Every time I wanted to try a new physics-informed neural network problem, I had to start almost from scratch: changing the PDE, updating boundary conditions, modifying the architecture,
