Applied AI Scientist Physics-grounded machine learning for defence and national security.
Theoretical Physicist Specialising in nonlinear and chiral light–matter interactions, and still maintaining scientific software in that field.
ORCID · Google Scholar · LinkedIn
Applied AI — evaluation, tooling, uncertainty
| Project | Summary |
|---|---|
| PhysBound | MCP server that lints RF and physical-layer calculations against hard physical limits (Shannon, Friis, radar range). Catches LLM physics hallucinations. On PyPI and the MCP Registry; CI + coverage. |
| RagOnAStick | Retrieval evaluation harness for UK MoD Joint Doctrine publications. Seven retriever/chunking configurations benchmarked (recall@k, MRR, nDCG) with MLflow logging. No generation — just measurement. |
| Chrono-Sentinel | Transformer-based time-series anomaly detection on the Numenta Anomaly Benchmark, with Monte Carlo Dropout uncertainty quantification and calibration analysis. |
Scientific software — simulation and physics-constrained ML
| Project | Summary |
|---|---|
| Sapphire | Post-processing environment for the structural characterisation of metallic nanoparticles and nanoalloys from MD trajectories — CNA signatures, coordination and aGCN, chemical ordering, change-point detection of melting transitions. Lead author; published in Faraday Discussions 242 (2023), pip install sapphire-nano, DOI-archived, CI + docs + executable tutorials. |
| Lumina | Rust framework for electromagnetic simulation of nanostructures via the Coupled Dipole Approximation. GPU-accelerated, O(N)-memory GMRES, Ewald-summed periodic systems, SHG/THG. 136 tests. |
| Metamaterials_PINN | Physics-informed neural networks for electromagnetic problems in metamaterials — Maxwell-constrained training with reproducible configs and tests. |
Also public: energy-demand forecasting across 10 US regions (LSTM/GRU/TFT vs gradient-boosted baselines), and unmaintained PhD-era code for HHG data processing, DFT tooling and nanoparticle dynamics.
- Applied AI Scientist, Whitespace (2026–present) — applied ML for defence and national-security problems.
- Data Scientist, The Alan Turing Institute (2025–2026) — ML and statistical modelling on HPC for AI research, and radio-frequency digital signal processing in defence and national-security contexts.
- Postdoctoral Research Associate, King's College London (2023–2025) — chiral and nonlinear light–matter interactions; computational and analytical frameworks for quantum-chemistry prediction.
Education: PhD Physics, KCL (2022) · MSc Non-Equilibrium Systems, KCL (2019) · MPhys Theoretical Physics, Leeds (2018)
I build ML systems that respect physical constraints: simulation, uncertainty quantification, and evaluation tooling that tells you when a model is wrong.
Languages: Python (PyTorch), Rust, C++, Fortran, SQL Scientific computing: MPI, CUDA, OpenMP, HPC schedulers Engineering: Docker, CI/CD (GitHub Actions), pytest, MLflow, packaging and release (PyPI), Linux, LaTeX