I build AI systems — RAG pipelines, evaluation frameworks, and applied ML tools for the public sector.
🚀 Currently at the Incubator for AI (DSIT) as programme lead for the Open Source AI Fellowship post-training foundation models for Government use cases on the UK's national AI supercomputer (Isambard), coordinating delivery with Meta and the Alan Turing Institute.
🌍 Previously at the Department for Energy Security & Net Zero, I designed and deployed production RAG systems serving multiple user groups.
stack = {
"llm_systems": ["RAG pipelines", "fine-tuning", "prompt engineering", "LLMOps"],
"frameworks": ["LangChain", "LlamaIndex", "Ragas", "FAISS"],
"cloud": ["Azure OpenAI", "Azure AI Search", "AWS EC2/S3", "HPC (Isambard)"],
"languages": ["Python", "SQL", "R", "bash"],
"practices": ["CI/CD", "evaluation frameworks", "responsible AI", "system design"],
}| Repo | What it does |
|---|---|
| govscan_streamlit | Scans and classifies international government GitHub organisations — live Streamlit app |
| name_origin | Probabilistic name-to-country inference with LLM-assisted ranking and evaluation harness |
PhD in Mathematics (University of Edinburgh) → Postdoc Research Fellow (Flatiron Institute, New York) → UK Civil Service AI roles. I've moved from large-scale stochastic modelling on HPC clusters to production LLM systems, but the underlying interest is the same: building things that work reliably in messy, real-world conditions.
I write about rebuilding technical skills in public on my Substack: covering RAG systems, fine-tuning, evaluation frameworks, and what it actually looks like to transition from strategic AI roles into hands-on engineering.
London · Warsaw · Remote