Data science, machine learning and AI agents — pointed at power. We map corporate networks, model the economy and read the documents nobody else has the capacity to.
LLMs changed the game: the scale of scraping, extraction and analysis we can do for a campaign budget is something only the biggest firms could attempt five years ago. We bring that capacity to the public-interest side.
Mapping power. We scrape filings, donations, contracts and the open web, then use LLMs to extract entities and relationships at scale — turning scattered records into a graph of who is connected to whom.
Microsimulation, input–output models and bespoke indices. We quantify the things policy debates assume but rarely measure — landlord profits, deprivation, the cost of climate shocks, the case for a shorter week.
LLM extraction over millions of documents — annual reports, job ads, regulatory filings, survey transcripts. We tag, classify and summarise unstructured text into datasets you can actually query.
We don't stop at the analysis. We ship public-facing interactive tools — searchable databases, trackers and indexes — so the people who need the findings can use them without a data team.
A slice of what we've built. The full back-catalogue runs to 60+ projects — reports, datasets, tools and demos.
Most of our work comes through people we've worked with before. If you've got a question that needs serious data capacity behind it, here's how we usually engage.
A discrete piece of research — scrape a corpus, build the dataset, run the model, deliver the findings (and the methodology) ready to publish or brief.
An interactive tool, dashboard or searchable database your audience can use directly — hosted, maintained and branded for you.
Large-scale searching and data aggregation for journalists — the heavy data lifting behind an investigation, on deadline.
Standing pipelines that keep a dataset fresh — donations, contracts, filings, risk signals — so you're never working from a snapshot.
We're researchers and engineers embedded in the Autonomy Institute — so we understand the politics of a brief, not just the parquet files.
Six data scientists and machine-learning engineers inside the Autonomy Institute.
Lead · ML & data engineering
Data science · ML research
NLP · network analysis
Investigations · political data
Economic modelling
Forecasting · statistics
Tell us roughly what you're trying to find out and the shape of the data. We'll come back with whether it's a fit and how we'd approach it.