Loading...
Loading...
Loading...
We automate public-source due diligence that traditional auditors often handle manually. Here's the technology behind it.
Track identity retrieval, synthesis, verification, scoring, and publication as separate recoverable stages.
Available provider and public evidence is canonicalized, deduplicated, and persisted before it can support a claim.
Every risk score is computed from typed, verifiable evidence — registry status, court findings, sanctions results — with a confidence rating, not a black-box guess.
When legal-risk evidence is available, each event records its entity match, status, date, materiality, and supporting evidence IDs.
Screening results are reported only when an available source supports the exact matched entity.
Our 0-100 risk score weighs 5 categories: business legitimacy (30%), legal compliance (25%), operational stability (20%), reputation (15%), and financial health (10%).
Available findings, evidence coverage, limitations, and next steps
How we decide when a report is complete, limited, or refunded
We run Gemini 2.0 Flash first, then optional Ollama Cloud specialist workers (researcher, China specialist, verifier) when configured. Strategy is recorded as gemini-ollama-parallel on each run.
Reports require at least five reliable sources from at least three independent origins before we mark research complete. Weak or contradictory evidence triggers a limited or failed outcome instead of guessing.
When evidence is too weak to publish a trustworthy report, we fail the run with a clear reason and restore the credit for that attempt.