I work on AI in finance. I did a PhD at UCD on machine learning, in particular learning embeddings for stocks from their returns time series. After that, I spent time as a research scientist at AWS, and now work on applied AI at Massive, where my recent work has involved turning unstructured financial documents into structured datasets.

I also tinker with a couple of side projects, and (sometimes) use this site to write up what I'm learning, mostly as notes for my future self.

Work

Massive, AI Research Lead 2025–

I built an SEC filings product, including LLM pipelines that tag risk factors in 10-Ks and events in 8-Ks against our own taxonomies. Before that, per-ticker news sentiment and a related-companies endpoint.

Amazon (AWS), Research Scientist 2022–25

Network availability. Worked out where in the region, zone and device hierarchy an outage was hurting. Two internships during my PhD before joining full time.

Projects

Polymarket market maker

Automated two-sided quoting on sports markets, loosely based on Avellaneda-Stoikov. Over $2M traded.

Research

Asset embeddings

My PhD. Learning vector representations of companies from which ones move together, similar to word embeddings in NLP. Useful for finding similar stocks, hedging and sector classification. ICAIF 2024 · blog post

LLMs for financial documents

Structured extraction from news and SEC filings, and how to evaluate it at scale. 10-K risks · 8-K events · news

Full list on Google Scholar.

Writing

• 2026-06Building a Parquet-backed data API, end to end
• 2026-04Digging Into Factor Modelling via Alternative Data
• 2026-04How to Add Login and a Paywall to a Streamlit App
• 2026-04Why Reading Parquet from S3 Is Slow (Even In-Network) and What I Did Instead
• 2026-01Optimising Query Transformations for Retrieval: A Negative Result
• All posts →