Vignesh Gopakumar
2024
I'm an AI researcher specialising in crafting "explainable" machine learning models that blend known physical laws with real-world data. Currently, I work as an AI Research Scientist at the UK Atomic Energy Authority, leading a team that builds actionable surrogate models for exascale simulations and data-driven models for plasma control and reactor design.
Additionally, I'm a visiting AI Researcher at the Rutherford Appleton Laboratory - STFC, where I contribute to foundational research in making machine learning models more robust and interpretable through physics-based approaches.
Current Research:
* Neural Operators
* Physics Informed Neural Networks
* Conformal Prediction
* Bayesian Optimisation
Building simvue.io : AI-driven, open-source simulation management and tracking dashboard for streamlining engineering workflows. Currently in private beta, it is being developed with public funding from the UK Government.
Surrogate Modelling
Crafting machine learning driven partial differential equation solvers.
Digital Twins
Real-time neural networks for forecasting diagnostic performance within experiments.
Design of Experiments
Uncertainty Quantification
Developing physics-informed and statistically valid uncertainty quantification tools for engineering applications.
Loss Landscape Engineering via Data Regulation on PINNs (Elsevier) - Vignesh Gopakumar, Stanislas Pamela and Debasmita Samaddar.
Fast Regression of Tritium Breeding Ratio in Fusion Reactors (IOP)- Petr Mánek, Graham Van Goffrier, Vignesh Gopakumar, Niko Nikolaou, Jon Shimwell and Ingo Waldmann
Image mapping the temporal evolution of edge characteristics in tokamaks using neural networks (IOP) - Vignesh Gopakumar and Debasmita Samaddar
Fourier Neural Operator for Plasma Modelling in Simulation and Experiments (IAEA FEC 23) - Vignesh Gopakumar, Stanislas Pamela, Daniel Brennand, Lorenzo Zanisi, Zongyi Li, Anima Anandkumar and Marc Deisenroth
Plasma Confinement Mode classification from Fast Camera Images (EPS 23) - Daniel Brennand, Vansh Tibrewal, Vignesh Gopakumar, Zongyi Li, and Anima Anandkumar
Evaluating imprecise probabilities in fusion plasma surrogates using conformal prediction (ISIPTA23) - Ander Gray, Vignesh Gopakumar, William Hornsby, James Buchanan, Stanislas Pamela
Multi-objective Bayesian optimisation for design of Pareto-optimal current drive profiles (SOFE 23) - Theo Brown, Stephen Marsden, Francis Casson, Vignesh Gopakumar, Alexander Terenin, Hong Ge
Fusion Reactor plant in-silico design and efficient simulation management case studies (SOFE 23) - Vignesh Gopakumar, Andrew Lahiff, Aby Abraham and Timothy Nunn
Scaling and Distribution of Physics-Informed Neural Networks for Fusion-Relevant Nonlinear Partial Differential Equations (ICDDPS Okinawa 23)- Lucy Harris, Vignesh Gopakumar and Stanislas Pamela
Fourier-RNNs for modelling noisy physics data (IEEE-ICMLA 22) - Vignesh Gopakumar, Stanislas Pamela, Lorenzo Zanisi
Fourier Neural Operator for Plasma Modelling (NeurIPS 22 Workshop) - Vignesh Gopakumar, Stanislas Pamela, Lorenzo Zanisi, Zongyi Li and Anima Anandkumar
Shaping of Magnetic Field Coils in Fusion reactors using Bayesian Optimisation (NeurIPS 22 Workshop) - Timothy Nunn, Vignesh Gopakumar, Sebastian Kahn
Towards real-time fusion reactor design using the Omniverse (Nvidia GTC 22) - Lee Margetts, Rob Akers, Abhijeet Ghosh, Vignesh Gopakumar, Patrik Hadorn, Mathias Hummel, Peter Messmer, Muhammad Omer, Ekin Ozturk, Stanislas Pamela, Pieter Peers, Benedict D. Rogers, Mark Rothwell, W I Sellers, Oliver Woolland
Active and continual learning of fusion plasma turbulence surrogate models for digital twinning of a tokamak device (ICML 21 Workshop) - Jackson Barr, Thandikire Madula, Lorenzo Zanisi, Vignesh Gopakumar, Aaron Ho, Jonathan Citrin, and JET Contributors
Informed Sampling of the Plasma Hyperspace for Digital Twinning (IAEA FDPVA 21) - Mayur Bakrania, Vignesh Gopakumar
14 MeV Neutron Irradiation Experiments - Gamma spectroscopy Analysis and Validation Automation (Physor 2020) - Thomas Stainer, Mark Gilbert, Lee Packer, Steven Lilley, Vignesh Gopakumar and Chris Wilson
FNO for Plasma Modelling (Invited) - IAEA Workshop on AI for Accelerating Fusion and Plasma Science (Vienna 2023)
FNO for Plasma Modelling - IAEA Fusion Energy Conference (London, 2023)
FNO for Plasma Modelling - AI for sustainability workshop @ UCL (London, 2023)
Fourier RNNs for modelling noisy physics data - IEEE ICMLA (Bahamas, 2022)
Informed Sampling of the Plasma Hyperspace for Digital Twinning - IAEA Fusion Data Processing, Validation, Analysis (Chengdu, 2021)
Optimising Physics Informed Neural Networks - PyTorch Ecosystem Day (Virtual, 2021)
Fluid Surrogates using Neural PDEs - SciML at RAL, STFC (Oxfordshire, 2020)
Solving Fluid Dynamics with Neural Networks - FusionEP (Virtual, 2020)
Data Driven Modelling of Plasma in Tokamaks - IOP Physics in the spotlight (London, 2019)
Data Driven Modelling and Control of Plasma in Fusion Reactors - O’Reilly AI London (London, 2019)
