Work
A living portfolio of things I've built, studied, or published. Updated as I ship more.
Research
- QTNM neutrino mass experiment
Machine learning + physics researcher, focussing on event detection and reconstruction.
Figuring out what the data means. Mostly it's noise. Occasionally, a neutrino. - Publications
Links out to arXiv / journal when they land.
Industry
- Protect Group · ML research intern, pricing
Risk-aware pricing on two years of refundable-booking data. A gradient-boosted profit model plus an inverse-propensity lift estimator for the counterfactual.
Raise prices on the risky half, leave the rest. Delivered as a cache-key pricing table. - Roku · speech ML intern, on-device ASR
Transformer-based speech recognition for 100M+ edge devices. Ten percent relative reduction in word error rate inside a tight latency and memory budget.
The TV hears you better, on-device, without phoning home. Groundwork for on-device conversational AI. - Peak AI (now UiPath) · AI research intern
Reinforcement learning (A2C, PPO) for inventory management, with custom sustainability penalties written into the reward.
The work seeded a sponsored PhD at Manchester.
Projects
- secret 1
A weekend rabbit hole in conservation zoology. Reading, noodling, counting things that would rather not be counted. Hobby scale. - secret 2
A long-running writing project on uncertainty and decisions. Slow cooked. A book one day, maybe. Nothing to ship. - nathanhigginbotham.com
This site. Astro + GitHub Pages + a single CSS file. Lean on purpose.
CV
No full PDF yet. The best route is to head to my LinkedIn and connect.