The Autonomy Data Unit is the in-house data science team at the Autonomy Institute. Six people, frontier methods, one job: turn filings, donations, contracts and the open web into evidence that holds up.
Each capability is a working pipeline, not a slide. We bring frontier tooling (large language models, supercomputer-scale processing) to questions that newsrooms, unions and campaigners actually need answered.
Scrape filings, donations, contracts and the open web. Use LLMs to pull out entities and the links between them. Then draw the map: who funds whom, who sits on which board, who keeps turning up.
Microsimulation, input-output models and bespoke indices. We build the number when the official one does not exist yet: landlord returns, job-risk scores, insecurity at work.
Point an LLM pipeline at millions of pages (company filings, policy documents, court records) and come out the other side with a clean, queryable dataset you can actually search.
Public-facing databases, trackers and indexes that outlive the report. Built to be used by reporters and organisers, not just downloaded once and forgotten.
Built since 2020 with unions, charities, newsrooms and campaigners. Live links where the work is public; some sits behind partners.

Millions of pages scraped to map the modern far right and how it plugs into money and power across the US and Europe.

An LLM pipeline that reads every UK annual report and surfaces confirmed risk events on a live map. Successor to our GERM system.

Political donations matched against government contracts. £138m in deals traced back to the firms that bankrolled the parties. Launched in the Guardian.

An AI-augmented index of the Heritage Foundation's 900-page plan, so anyone can search what the manifesto actually proposes.

Labour's tilt toward business donors, tracked from 2019 to 2024. Big-business giving rose roughly 1,700% across the period.

An economic model of UK landlord returns for the Joseph Rowntree Foundation. The number nobody else had built.

A searchable database of licensed care-visa sponsors, built with the Bureau of Investigative Journalism for reporters working the care beat.
Arts Council funding broken down by constituency since 2014, made browsable for Equity and its members.

The corporate connections behind the UK's entrepreneurial far right, mapped from filings and the open record.

The origin file. UK workers scored by Covid exposure: nearly 11 million in higher-risk roles. Featured on ITV's Peston in April 2020.

30 million job ads tagged with LLMs on the Isambard supercomputer to measure AI exposure across the labour market, with the UK AI Security Institute.

A co-mention network built for the International Trade Union Confederation to trace the companies working against workers' rights worldwide.
A small team of data scientists and machine-learning engineers. We came out of frontier ML and pointed it somewhere useful. Part of the Autonomy Institute.
We work with unions, charities, newsrooms and campaigners. Tell us what you're trying to prove and we'll tell you whether the data can carry it.
adu@autonomy.work →