Skip to content
Gyanateet Dutta
Work·Academic·CV

Work

Systems

Physics-Informed Neural Networks

This repository contains GPU implementations of PDE solvers and physics-informed neural networks for Navier–Stokes, Burgers, and related nonlinear equations, using CUDA C++ and PyTorch. The picture is the Burgers benchmark solved exactly, not output from these notebooks.

Forty-four stacked ink curves of Burgers’ equation, from a smooth sine wave at t = 0 to a sharp shock at x = 0 by t = 1.
Fig. 1Burgers’ equation with ν = 0.01/π, drawn one time slice at a time: the sine wave steepens into a shock at x = 0. Exact solution by the Cole–Hopf transform, computed for this figure. It is the benchmark PINNs are fitted to, not output from the repository’s networks.
Scope

Method-exploration notebooks; validation against CFD reference solvers is future work.

Stack
CUDA, PINNs, CFD, Scientific Computing
Links
Code

Also

  • 2026CUDA Blackwell Labs
  • 2026Agentic structure-from-motion
  • 2026PCOS edge agent
  • 2024Dalton Mills VR Reconstruction
  • 2023AWS AI/ML Scholar
GitHub· ORCID· Google Scholar· LinkedIn· Hugging Face· X