Ben Whewell
AI/ML Scientist at Los Alamos National Laboratory
I work on neutron transport, nuclear data, and machine learning. Most of my research is about making deterministic transport solvers faster and more accurate: reinforcement learning for multigroup energy grid optimization, hybrid numerical methods, and solution verification. I also build the ML tooling around that work, from training pipelines for simulation data to retrieval-augmented generation (RAG) tools for technical documents.
I finished my PhD in Mechanical and Aerospace Engineering at the University of Notre Dame in January 2025. Before graduate school I spent three years as a boilers and turbines operator at a power station in Connecticut.
Research
-
Reinforcement learning for energy grids
Training RL agents to choose multigroup energy structures for criticality problems.
-
Collision-based hybrid method
Splitting the transport solve into uncollided and collided parts that use different resolutions.
-
Solution verification
Estimating discretization error in transport solutions without running a mesh refinement study.
-
ML for nuclear data reduction
Replacing large scattering and fission matrices with neural network surrogates.
News
- May 2026
- Started as an AI/ML Scientist at Los Alamos National Laboratory.
- May 2026
- Posted a preprint on reinforcement learning for multigroup energy grid optimization (arXiv:2605.27895).
- 2025
- Received a conference paper award at the 2025 ANS Mathematics and Computational Methods conference (M&C 2025).
- Feb 2025
- Started the Metropolis Postdoctoral Fellowship at Los Alamos National Laboratory.
- Jan 2025
- Finished my PhD at the University of Notre Dame (dissertation).