ADU/TERM v2.0 — SESSION OPEN PUBLIC-INTEREST DATA SCIENCE LEDGER REC ● 2020—2026
Autonomy Data Unit / Internal Reference Document
RFC ADU-0001
FILE NO. 2020—2026
REV. 2026.06

Data science,
pointed at power.

We are the Autonomy Data Unit, the research and engineering arm of the Autonomy Institute. Six data scientists and machine-learning engineers. Think of us as the data arm of the public interest. Since 2020 we have built the trackers, indexes and investigations that unions, charities, newsrooms and campaigners use to hold power to account. Frontier methods, movement values.

6 people 2020 founded LLMs at supercomputer scale network · economic · document intelligence
§01

Capabilities

FOUR LINES OF WORK
01

Network analysis

We scrape filings, donations, contracts and the open web, then run LLM entity and link extraction to map who is connected to whom. The output is a graph you can interrogate.

METHOD: graph build
INPUT: filings, donations
SEE: Six Degrees of Reform
02

Economic & labour modelling

Microsimulation, input-output models and bespoke indices. We turn messy ONS and administrative data into a number that means something and can stand up to scrutiny.

METHOD: microsim, I-O
INPUT: ONS, admin data
SEE: The Property Premium
03

Document intelligence

LLM extraction across millions of documents, collapsed into a single queryable dataset. Annual reports, planning records, policy plans: read once, search forever.

METHOD: LLM extraction
SCALE: millions of docs
SEE: Risks to British Business
04

Tools & data products

Public-facing searchable databases, trackers and indexes. The kind of thing a journalist can use at deadline and a campaigner can cite in a hearing.

METHOD: web app + DB
OUTPUT: live, public
SEE: Care Visa Database
§02

Selected ledger of work

12 ENTRIES / SELECTED
REF
PROJECT
YEAR
STATUS
A-01
The Authoritarian Stack
The Authoritarian Stack
Millions of pages scraped to map the modern far right and its links to power.
network · report
2025
LIVE ►
authoritarian-stack.info
A-02
Risks to British Business
Risks to British Business
An LLM pipeline reading every UK annual report to surface confirmed risk events.
document intel · live
2026
LIVE ►
riskstobritishbusiness.today
A-03
AI Exposure Index
30M Job Adverts / AI Exposure
30 million job ads tagged with LLMs on the Isambard supercomputer, with the UK AI Security Institute.
document intel · AISI
2025
NO PUBLIC URL
internal / AISI
A-04
Givers and Takers
Givers & Takers
Political donations linked to government contracts. Launched in the Guardian.
network · report
2025
LIVE ►
autonomy.work
A-05
Labour, the Party of Capital?
Labour, the Party of Capital?
Labour's shift toward business donors, traced from 2019 to 2024.
network · report
2026
LIVE ►
autonomy.work
A-06
Project 2025 Index
Project 2025 Index
An AI-augmented index of the Heritage Foundation's 900-page plan.
document intel · tool
2024
LIVE ►
project2025index.com
A-07
The Property Premium
The Property Premium
An economic model of UK landlord returns, built for the Joseph Rowntree Foundation.
economic model · JRF
2025
LIVE ►
autonomy.work
A-08
Care Visa Sponsorship Database
Care Visa Sponsorship Database
A searchable database of licensed care-visa sponsors, built with the Bureau of Investigative Journalism.
tool · TBIJ
2024
LIVE ►
autonomy.work
A-09
Arts Funding Tracker
Arts Funding Tracker
Arts-council funding by constituency since 2014, built for Equity.
tool · Equity
2024
LIVE ►
apps.autonomy.work
A-10
Six Degrees of Reform
Six Degrees of Reform
Mapping the corporate connections of the UK's entrepreneurial far right.
network · report
2024
LIVE ►
autonomy.work
A-11
Corporate Underminers
Corporate Underminers
A co-mention network built for the ITUC, the global trade-union body.
network · ITUC
2025
NO PUBLIC URL
ITUC / internal
A-12
Jobs at Risk Index
Jobs at Risk Index
The origin project. UK workforce scored by Covid exposure, featured on Peston in 2020.
tool · origin
2020
LIVE ►
autonomy.work
§03

Personnel

A SMALL TEAM INSIDE THE AUTONOMY INSTITUTE
EMP-01 / LEAD
Lukas Kikuchi
Unit lead. Machine learning and data engineering. Built JARI in 2020, runs the pipelines now.
EMP-02
Bhargav Srinivasa Desikan
Data science and ML research. Frontier methods, applied to public-interest problems.
EMP-03
Sean Greaves
NLP and network analysis. Turns documents and filings into graphs you can read.
EMP-04
Sonia Balagopalan
Investigations and political data. Lead author on the donations and contracts work.
EMP-05
Luiz Garcia
Economic modelling. Microsimulation, input-output models, and the indices that anchor reports.
EMP-06
Jeremy Kwok
Forecasting and statistics. Keeps the numbers honest and the uncertainty visible.
§04

Worked with

PARTNERS & COMMISSIONERS
  • AI Security Institute
  • Joseph Rowntree Foundation
  • Joseph Rowntree Reform Trust
  • Equity
  • Unite the Union
  • ITUC
  • Good Law Project
  • Bureau of Investigative Journalism
  • Centre for Investigative Journalism
  • Future Economy Scotland
  • Spotlight on Corruption
  • TfL / GLA
§05

Open a line

END OF DOCUMENT
> TO COMMISSION WORK, WRITE TO:
adu@autonomy.work

We win most of our work through people we already know. If you run a union, a charity, a newsroom or a campaign and there is a dataset you wish existed, send a line. We will tell you straight whether we can build it.