I'm a Mathematics and Statistics student at the University of Warwick. I like problems where a model has to survive contact with messy data: limit order books, options quotes, electricity markets, and chip layouts. Most of my work is in Python, with C++ when I need tighter control over a model or simulation.
Right now I'm building an options-pricing and implied-volatility engine for Elphinstone Research Group. I'm also working in a 10-person Warwick Quant team on Atlas, where we turn public energy, market, company-filing, and grid-connection data into an inspectable research report.
I enter competitions fairly often. A deadline is a good way to find out whether an idea actually works.
| Project | What I did |
|---|---|
| WunderFund LOB Predictorium | Published my exact v96.1 submission, combining GRU, LSTM and temporal-convolution models with ONNX CPU inference. Reported competition result: 49th out of 5,000, with a 0.3098 weighted-correlation score. Includes model artifacts, a synthetic demo, evaluation tooling and 23 automated tests. |
| Atlas | Built with a 10-person Warwick Quant team. The research pipeline processed 529,434 electricity, wholesale-price and company-filing observations across seven US regions, with interconnection evidence presented separately. Includes source-linked reports, 64 tests and 86% package coverage. Weak historical associations and data-vintage limitations are documented; this is a descriptive tool, not an alpha model. |
| K2HAL | Placed 26th out of 200 in the HRT x Partcl challenge. Built a CPU-only chip-layout search with an incremental scorer; the submission write-up reports a 1,800x scoring speedup under its benchmark conditions. Returned zero-overlap layouts on all 17 IBM benchmarks, with committed result evidence and baseline comparisons. |
| M3 Challenge — housing and homelessness | Led a five-person team during the 2024 competition. Published the original 29-page team report and code screenshots, alongside a clearly labelled 2026 Python reconstruction of selected regressions. Includes 20 tests, 98% coverage and explicit limitations on long-range extrapolation. |
| Universal Options Engine (private research tool) | Building a Python pricing and implied-volatility engine for Elphinstone, with quote-quality checks, reproducible command-line workflows, and 22 tests. |