Portfolio · LinkedIn · Substack · X
Hi, I'm Padraig.
I'm an AI engineer, currently doing an MSc in Artificial Intelligence at the University of Edinburgh. Before that I spent three years at yieldHUB, taking ambiguous product problems through to systems customers actually used — LLM, RAG and computer-vision features that shipped inside its semiconductor analytics platform.
I started in data science, but found myself increasingly interested in what happened after the model worked: does it still hold up on real data, and would anyone notice if it stopped? I taught myself to build software properly so I could answer that, and I've done it in public, mistakes included.
Getting an answer out of a model is the easy part. The harder part is knowing whether the answer is any good — evals, regression gates, traces, failure analysis. It's most of what I work on now.
If a system can't say "I don't know", I don't trust it. Every repo pinned below has a file listing what it can't do. When I look at someone else's project, that's the first thing I read.
- Doing an MSc in Artificial Intelligence at the University of Edinburgh
- Working mostly in Python / TypeScript — FastAPI, Next.js, Postgres/pgvector, PyTorch
- nanogpt-pob.dev — drag a scrubber along a GPT-2 reproduction and watch it cross a target that was written down before the run
- pob-rag-chat.xyz — ask a question, then follow the answer back through retrieval, cost and latency
- strideiq-lemon.vercel.app — a year of sample training data, analysed in your browser; no account, no key, nothing to install
I built StrideIQ because I wanted to know why my training had gone backwards. It told me. That's roughly the pattern for everything here.
Looking for AI / ML engineering roles.
Based in Edinburgh for the MSc. Open to London, the US and elsewhere.
If you're building AI systems where reliability matters, I'd like to hear about it.