The platform
Three products on one governed foundation.
Every request follows the same governed path: the data rules are set first, the work is planned, sources are retrieved and numbers are computed by tools, every claim is verified, and the whole chain is recorded.
Sparse MoE economics. Globally representative data. Audit-ready compliance.
Built on the sparse Mixture-of-Experts paradigm and hardened for regulated life sciences: BioMistral-7B base, multi-LoRA expert specialisation, the GenoNet knowledge graph, multimodal RAG and a GxP-validated promotion gate.
- 128Kcontext window: full protocols and evidence chains in one prompt
- 2.4×faster generation via multi-token prediction
- 85%lower compute per query from sparse Top-K activation
- 100%of outputs carry citations and provenance
The data moat that matters to your trials
- 1,000curated datasets
- 97.5%coverable via free or free-substitute paths
- 120K+Indian, African and Caucasian oncology records
- 4-tierlicence, PHI and drift confidence per field
The rules come first. The answer comes last.
- 1 · PolicySet the data rulesBefore any cache, retrieval or model call
- 2 · PlanClassify and routeOut-of-scope requests stop here
- 3 · RetrieveSources and toolsDocuments, knowledge graph, vocabularies
- 4 · ComputeNamed calculatorsThe model never writes a number
- 5 · VerifyCitation gateSpan match and entailment per claim
- 6 · RecordAudit and reviewHash-chained record; conflicts go to a person
Simplified from the eleven-step request lifecycle.
GenoDev · Clinical Development
A trial design your biostatistician can check.
The model selects and cites the assumptions. A deterministic tool computes the sample size. The screen shows both, and editing any input recomputes the result.
- AI protocol authoring aligned to ICH M11
- In-silico trial simulation and site scoring
- Biomarker and genomic integration
- Adaptive design and feasibility forecasting
- EHR and claims-driven patient recruitment
For clinical development leads, medical directors and biostatisticians
GenoRWE · Real-World Evidence
Know how many patients are really there, and why the rest aren’t.
Criteria resolve to standard concept sets. Counts come from direct queries against the data, with no model involved. Every step of the funnel links to the concept set that produced it.
- OMOP CDM-harmonised federated search
- Rare-disease prevalence and burden of illness
- Comparative effectiveness
- Claims, EHR and registry triangulation
- Label expansion and post-market RWE
For feasibility, clinical operations, RWE and epidemiology leads
GenoAccess · Market Access & Value
Payer evidence built from the same population as the trial.
The eligible count from GenoRWE drives the budget impact directly. When assessment bodies or studies disagree, both are shown with their design labelled, and the artifact is held for review.
- HEOR cost-effectiveness modelling
- Payer policy and pricing simulation
- Auto-drafted value dossiers and AMCP packages
- Competitive intelligence and HCP segmentation
- Outcomes-based contract design support
For market access, HEOR and pricing leads
Every number has a method and a version.
Arithmetic is never left to a language model. Each computed figure carries its method and tool version into the audit record, so a reviewer can reproduce it.
| Product | Calculation | Method |
|---|---|---|
| GenoDev | Events and sample size, time-to-event | Schoenfeld, with Freedman shown as an alternative |
| GenoDev | Enrolment projection | Sites × rate × duration, accrual model named |
| GenoRWE | Attrition and eligible population | Direct query, no model involvement |
| GenoRWE | Time-to-event summary | Kaplan–Meier, censoring rule stated |
| GenoAccess | Cost-effectiveness ratio | Incremental cost ÷ incremental QALYs |
| GenoAccess | Budget impact and discounting | Eligible population from GenoRWE; discount rate per assessment body |
Outputs that look like they came from someone who has done the work.
- Protocols
- ICH M11 structure
- Cohorts
- OHDSI and OMOP conventions
- Observational outlines
- STROBE
- Economic evaluation
- CHEERS 2022
- Evidence grids
- PICO rows, study design on every row
Talks to the stack you already run.
- Data in
- OMOP CDM · HL7 FHIR R4 · SDTM/ADaM · DICOM · CDISC ODM · S3 · SFTP
- Identity
- SAML 2.0 · OIDC · SSO · MFA · per-tenant RBAC
- API surface
- REST · GraphQL · OpenAPI 3.1 · gRPC · webhooks
- Workflow
- Veeva Vault · Medidata Rave · Oracle Argus · Epic / Cerner / OpenMRS
Three deployment modes. Switch as you scale.
Pick the mode that fits your data residency, IT and risk posture. No re-onboarding when you move.
- Mode A
Managed SaaS
Genovant-hosted on AWS with per-tenant logical isolation. The default for fast pilots.
Best for: Design-partner pilots, mid-size customers, biotech, regional CROs
- Mode B
Customer VPC
Deployed inside your AWS, Azure or GCP account, with keys managed by you.
Best for: Top-20 pharma, large CROs, regulated data planes, GxP production workloads
- Mode C
On-premises / air-gapped
Container-based delivery into your data centre, for sovereign or classified data.
Best for: Government health, sovereign-residency mandates, classified trials
From kickoff to first usable output
- 2 wksdata on-ramp and sandbox
- 4 wksfirst usable model output
- 8 wkspaid pilot complete
- 6–12 moenterprise general availability