Seven years of proprietary diligence data, made searchable and predictive by a purpose-built graph neural network.
21,000,000 patents and papers were filed in 2025. The most valuable technical fields are now effectively unsearchable. Huge advances in LLMs have allowed for sampling and interpretation, but expert input still represents a bottleneck.
Manual diligence on a single company can take weeks of specialist time, and requires multiple experts to minimise conflicts.
Searches on technical search platforms often return 1,000,000+ results — in our projects, ~93% aren't relevant to the client. Shell companies, conflicting IP and stealth competitors stay invisible to web search. 30–60% of scientists aren't on networks like LinkedIn.
Generic intelligence platforms cost ~$20k and still leave the analyst to do the reading.
Left: raw co-authorship graph. Right: the same graph after Erasmus isolates the relevant cluster.
Every paid project adds labelled reports and outcomes to a knowledge base no competitor can replicate.
Insight from past searches makes each new search faster and sharper.
Outcome labels start to turn the graph into a predictive model, not just a map.
Models improve with each use as the engine does more of the work experts used to do by hand.
40+ paid studies across VC, biotech, tech, public institutions and corporate venture — over 5,000,000 documents analysed.
504 key scientists and 40 new companies surfaced from 25 years of research — delivered in under 5 days.
95,000 skilled people mapped, 20,000 in target geographies — supported hiring across 200+ roles.
300,000 documents across 3 European countries → 19,661 key authors and 6 biotech competency areas, in 2 days.
Has found multiple stealth companies and IP holding companies on real client projects.
2-week accelerated engagements. Transparent pricing on analysis + storage (~£0.10/data point) + expert input.
Recurring, secure online access to results and tooling. Priority on repeat and vigilance projects.
Run your own searches, vigiliance and monitoring at scale
We support you on your data science journey with a range of flexible ways of working.
Proprietary diligence data, a graph neural network already in production, and a client base that keeps feeding both.
350+ reports, 40+ paid pilots, 5,000,000+ data points analysed on real client work.
Seven years of labelled outcomes and a GNN already in production — not a research idea.
Every project makes the graph richer, the predictions better, and the margins wider.