About

I spent four years learning how physical systems fail, then decided I was more interested in the data coming off them than the systems themselves. That's the short version of how a mechanical engineering degree from Loughborough turned into an MSc in Data Science and AI at Liverpool.

The MSc finished with a distinction and a dissertation on forecasting S&P 500 volatility — LSTMs against the GARCH family, plus a hybrid that I was certain would win and didn't. That result is written up here, negative finding and all, because being able to explain why something failed is worth more than a clean chart.

I was the technical hire at a small data and automations consultancy — dashboards, pipeline instrumentation, reporting that non-technical people actually read. Most of it wasn't glamorous. It was finding out why a number was wrong, tracing it to a form nobody filled in properly, and fixing the form rather than the spreadsheet.

Outside of it: semi-pro football for a while, a Nike Run Club competition I entered late with no running shoes and won by an embarrassing margin, and a newsletter I abandoned the moment people started reading it. I'm working on that last one.

• MSc Data Science & Artificial Intelligence
  University of Liverpool — Distinction 2025 — 2026
• BEng Mechanical Engineering
  Loughborough University 2018 — 2022

Split by what I'd be comfortable being tested on, rather than by what looks good in a list.

Confident

• Python
  Everything I build starts here. pandas and NumPy for the data, Matplotlib when I need to see it.
• PyTorch
  Every LSTM in my dissertation. I picked it over TensorFlow because the training loop is explicit — you can see what's happening on each step, which mattered when I was learning.
• scikit-learn
  Baselines and preprocessing. The first thing I reach for before anything more complicated, because a linear model you understand beats a deep one you don't.
• Git
  Branches and pull requests as a matter of course, including on this site.

Working knowledge

• SQL
  I can read it and write straightforward queries. I'm actively working on this — it's the biggest gap between what I can do and what I want to do.
• Supabase
  Where the data sits in my current work. Postgres underneath, so most of what I learn here transfers anywhere.
• n8n
  Automation and pipeline glue. Useful mainly because it makes the shape of a broken process obvious.
• HTML & CSS
  Enough to build this site by hand, which is why it has no framework in it.