The platform
One engine. Three ways to use it.
Commission an independent review, run analyses with the checks built in, or — from mid-2027 — call the same checks from your own pipeline or agent. Same rules, same evidence, whoever runs it.
For investors & BD teams
Diligence
Independent review of the computational claims behind a financing, licensing deal or trial decision — a written verdict you can act on, backed by signed records you can check.
- Claim inventory mapped to the analyses behind each claim
- Independent re-analysis of every examined claim
- Claim-by-claim verdict report with structured findings
- Signed records for every analysis we run
For research teams
Runtime
Supported analyses with the checks built in. Every run produces a signed record — ready for a reviewer, a partner or a data room.
- Differential expression (RNA-seq counts, microarrays, single-cell pseudobulk), enrichment, clustering and ML
- Checks run before execution: confounding, pseudoreplication, data type, plan deviations
- Signed analysis records; RO-Crate and W3C PROV export; TRIPOD+AI, REMARK and STARD-AI checklists
- Portable evidence packages and scheduled replay — in development, 2027
For AI-science platforms · developer preview mid-2027
Assurance API
The external system orchestrates; Genomarker checks, records and signs. The same rules and the same evidence, whoever the actor is.
- Schema discovery — agents learn what can be run and how
- Request validation before execution; blocking findings returned machine-readable
- Idempotent submits, safe to retry and resume
- The actor — human, pipeline or agent — recorded in every record
Under every surface
The same lifecycle, the same evidence.
- Ingest
Content-address every input. Identity before analysis.
- Pre-execution
The checks evaluate the request. Analyses the data can't support are refused before they run.
- Execution
The requested computation runs exactly as specified.
- Evidence
Findings, provenance, environment and results — recorded and signed.
- Preservation
Every replay dependency classified and retained — in development, mid-2027.
- Replay
An isolated environment re-executes the analysis and attests to the outcome — in development, mid-2027.
The reasoning behind each check lives in how it works — Read how it works.
Early access, in small cohorts.
We're onboarding research teams, AI-science platforms and review clients now. A person replies within 5 business days.