We are the Autonomy Data Unit: six data scientists and ML engineers inside the Autonomy Institute. Since 2020 we have built the evidence for unions, charities, newsrooms and campaigners using the same frontier methods the other side keeps to itself.
Most jobs we take on are some mix of these. The methods are current. The questions are old: who holds power, how, and at whose expense.
We scrape filings, donation registers, contract awards and the open web, then use LLMs to pull out the people and organisations and the links between them. The output is a map of who is connected to whom.
Microsimulation, input-output models and bespoke indices. We turn messy public statistics into a number a campaign can stand behind and a journalist can quote.
LLM pipelines that read millions of documents (annual reports, policy plans, filings) and fold them into one queryable dataset. We did this on a supercomputer when the document count ran to the tens of millions.
When the finding needs to outlive the report, we ship it: searchable databases, trackers and indexes that a non-technical team can keep using long after we have left.
A dozen of the things we have shipped, for clients including the AI Security Institute, the Joseph Rowntree Foundation, Equity, the ITUC and the Bureau of Investigative Journalism.
Six people. No account managers, no middle layer. You work with the person writing the code. We sit inside the Autonomy Institute, an independent think tank for the future of work and the economy.
Most of our work starts with a conversation, often through someone we already know. Tell us what you are trying to prove, or disprove, and we will tell you straight whether the data is there.
adu@autonomy.work
Data Unit