Vol. VI  ·  No. 1  ·  Established 2020 Public-interest data science  ·  London & remote Price: one introduction
The
Data Unit
Dispatches from the Autonomy Data Unit A wing of the Autonomy Institute Since 2020

Data science, pointed at power.

Six data scientists and machine-learning engineers, working as the research and development arm of the Autonomy Institute. We scrape the filings nobody reads, run language models across millions of documents, and turn the result into evidence that unions, charities, newsrooms and campaigners can act on. The joke we tell about ourselves is that we left Palantir for the other side. The serious version: this is the data arm of the public interest. We have been doing it since 2020, starting the week the pandemic hit, and we have not stopped.

What we do

№ 01

Network analysis

We scrape filings, donations, contracts and the open web, then use language models to pull out entities and the links between them. The output is a map of who is connected to whom, and what changed hands.

№ 02

Economic & labour modelling

Microsimulation, input-output models, and bespoke indices. When the existing statistics do not answer the question, we build the measure that does and show our working.

№ 03

Document intelligence

LLM extraction across millions of pages: annual reports, registers, planning files. We turn unstructured text into a dataset you can query, with the source kept attached to every fact.

№ 04

Tools & data products

Public-facing searchable databases, trackers and indexes. The work does not stop at a PDF. We ship things people can use, and we keep them running.

The front page

The Authoritarian Stack
Plate I. Millions of pages, scraped and linked. The Authoritarian Stack, 2025.
Investigation · Network analysis · 2025

The Authoritarian Stack

We scraped millions of pages to map the modern far right and trace its connections to money and power across America and Europe. Built with the Autonomy Institute, published as a standalone site that reads like a wiring diagram of the post-democratic project.

Read the investigation →

Dispatches

Risks to British Business
Plate II. Every UK annual report, read by machine.
Tool · Document intelligence · 2026

Risks to British Business

An LLM pipeline reads every UK annual report and surfaces the risk events companies have actually confirmed. Live, and updating.

Open the monitor →
AI Exposure Index
Plate III. 30 million job ads, tagged on a supercomputer.
Index · Labour modelling · 2025

30M Job Adverts, Tagged by AI

Thirty million job ads scored for AI exposure using LLMs on the Isambard supercomputer, built with the UK AI Security Institute.

Built with AISI · no public site
Givers and Takers
Plate IV. Donations, traced to contracts.
Investigation · Network analysis · 2025

Givers & Takers

Political donations linked to government contracts. Launched in the Guardian; the donor-contractor nexus, named and counted.

Read the report →
Labour, the Party of Capital?
Plate V. Where the money moved, 2019 to 2024.
Investigation · Political data · 2026

Labour, the Party of Capital?

Labour's shift toward business donors between 2019 and 2024, drawn straight from the donations record. Picked up by Novara.

Read the report →
Project 2025 Index
Plate VI. 900 pages, made searchable.
Tool · Document intelligence · 2024

Project 2025 Index

An AI-augmented index of the Heritage Foundation's 900-page plan, so anyone can find what it actually says.

Search the index →
The Property Premium
Plate VII. An economic model of landlord returns.
Report · Economic modelling · 2025

The Property Premium

An economic model of landlord profits across England, built for the Joseph Rowntree Foundation. The numbers behind the rent.

Read the report →
Care Visa Sponsorship Database
Plate VIII. A register, made useful to the people in it.
Tool · Data product · 2024

Care Visa Sponsorship Database

A searchable database of licensed care-visa sponsors, built with the Bureau of Investigative Journalism for the workers who need it.

Open the database →
Arts Funding Tracker
Plate IX. Arts funding, constituency by constituency.
Tool · Data product · 2024

Arts Funding Tracker

Arts-council funding by constituency since 2014, built for Equity so members can see where the money went.

Open the tracker →
Six Degrees of Reform
Plate X. The corporate map behind the entrepreneurial far right.
Investigation · Network analysis · 2024

Six Degrees of Reform

Mapping the corporate connections of the UK's entrepreneurial far right, one directorship at a time.

Read the investigation →
Corporate Underminers
Plate XI. A co-mention network for the global labour movement.
Investigation · Network analysis · 2025

Corporate Underminers

A co-mention network built for the International Trade Union Confederation, tracking the companies working against organised labour worldwide.

Built with the ITUC · no public site
Jobs at Risk Index
Plate XII. The origin file, March 2020.
Tool · Labour modelling · 2020

Jobs at Risk Index

Where it started: the UK workforce scored by Covid exposure, on ITV's Peston within a fortnight of lockdown.

Read the original →

The masthead

A small team inside the Autonomy Institute.

Six people. No account managers, no layers. You talk to the person doing the work, and the work is built to be checked. We take on projects for unions, charities, newsrooms and campaigners, and most of our work arrives through someone we already know.

If that is you, the email is at the bottom of the page.

Lukas Kikuchi
Lead · ML & data engineering
Bhargav Srinivasa Desikan
Data science · ML research
Sean Greaves
NLP · network analysis
Sonia Balagopalan
Investigations · political data
Luiz Garcia
Economic modelling
Jeremy Kwok
Forecasting · statistics

Who we work for

AI Security Institute Joseph Rowntree Foundation Joseph Rowntree Reform Trust Equity Unite the Union ITUC Good Law Project The Bureau of Investigative Journalism Centre for Investigative Journalism Future Economy Scotland Spotlight on Corruption TfL / GLA