In detail
- Raised $9M seed funding from Andreessen Horowitz.
- First product: a data‑science tool that returns answers with citations and an audit trail and validates results against deterministic datasets.
- Architecture trains models against a validator and uses a harness so the system can run on significantly weaker models, potentially on local hardware.
Why it matters
Businesses need dependable, auditable AI outputs; an approach that reduces hallucinations and enables smaller, local models can lower both risk and operating cost.
For you Evaluate how output validation and audit trails are handled in your AI vendors; test validator‑based workflows to reduce hallucinations and token costs.