Shahid M.

I build physics-informed neural networks (PINNs) and DeepONet models that actually respect your physics, trained and validated against analytical solutions, numerical results or measured data. Recent result: a DeepONet surrogate for an elastoplastic mechanics problem, validated on 500 test cases at 0.43% mean and 0.52% maximum relative error, with formal statistical testing. I'm a physics PhD candidate (computational astrophysics, NUST) and work daily with PDEs, numerical methods and simulation, so I read your equations, not just your dataset. WHAT I DO - PINNs and DeepONet for forward and parametric PDE problems; inverse problems scoped case by case (DeepXDE, PyTorch, TensorFlow) - Validation: error metrics and comparison with analytical, FEM or FDM solutions - Numerical simulation and parameter studies in Python, C++ and Fortran - Publication-quality figures in Python/Matplotlib, sized for your journal ALSO AVAILABLE - LaTeX/Overleaf: journal and conference templates (IEEE, Elsevier, Springer), Word-to-LaTeX conversion, compile fixes - Scientific accuracy review of physics and STEM content: books and course material HOW I WORK I work to a fixed scope with clear milestones. Before you hire me, I'll tell you whether a PINN is the right tool for your problem, or whether a classical solver would serve you better. Send your equations, boundary conditions and any reference data, and I'll reply with a plan.