Clinical-grade synthetic data — with proof
Gesalp AI turns sensitive clinical tables into synthesist real data: realistic, privacy-safe datasets audited for utility and compliance. Built for sponsors, CROs, and AI teams.
HIPAA/GDPR-ready. On-prem or SaaS.
Agent data scientist
Our agent selects the right generator (GC/CTGAN/TVAE/diffusion), tunes hyper-params, and optimizes for your objective (privacy-first, utility-first, or balanced).
Reproducible seeds & run cards.
Privacy with proof
Every dataset ships with an audit: DCR p01, MIA advantage, k-anonymity, and optional DP (ε, δ). Green-light gates block release until thresholds pass.
Exportable PDF & JSON.
Trial-aware synthesis
Survival/censoring aware. Supports longitudinal vitals/labs and irregular time-series.
Tabular today; diffusion & transformers next.
Deploy anywhere
Use our SaaS or run on-prem. The on-prem engine only publishes when all privacy/utility gates are green.
Governance & lineage
Dataset cards capture source URI+hash, cohort ε, metrics, and post-processing steps for audit trails.
Marketplace ready
Publish to your catalog: consumers can use the data without owning the originals—Snowflake-style access controls.
How it works
Upload or connect
Securely upload CSV/Parquet or connect your table.
Click Start
Agent chooses model + tuning; optional advanced settings.
Evaluate
See privacy & utility metrics update in real time.
Deliver
Download CSV/Parquet + audit PDF/JSON, or publish to marketplace.
How we'll maintain and extend our product advantage
Assess data sources
Inventory OMOP/FHIR/CSV and define cohorts.
Prepare sample cohort
De-risk schema, units, code sets (ICD/ATC/LOINC).
Train DP engine
Auto-select model (TabDDPM/CTGAN/TVAE) with ε/δ budget.
Run privacy audits
DCR/kNN, MIA advantage, k-anon/ℓ-div; gate on pass.
Validate utility
TSTR/RTS, AUROC/AUPRC vs real; drift (KS/AD, PSI).
Generate dataset card
Lineage, ε/δ, audits, hash; PDF + JSON.
Pilot with partners
Hospital/CRO runs on-prem/VPC; collect feedback.
Publish & monitor
Marketplace listing, usage telemetry, retrain loop.
Why Gesalp
Measurable privacy
Distance-to-closest-record, membership inference AUC, k-anon/ℓ-diversity, optional DP accounting.
Real utility
TSTR/RTS, AUROC/AUPRC, drift (KS/AD/PSI), and correlation/MI delta—benchmarked vs. real baseline.
Enterprise-ready
On-prem mode, SSO, role-based access, air-gapped deployments, and regulator-friendly reports.
Pricing
GESALP AI supports teams of all sizes, with pricing that scales.
Research
Perfect for academic research
- 1 project
- ≤5k rows
- GC/CTGAN
- Basic privacy–utility report
Starter
For growing teams
- 5 projects
- GC/CTGAN/TVAE
- Full utility suite
- Email support
Pharma/Enterprise
For enterprise needs
- Unlimited projects
- Diffusion + DP
- Full privacy suite + audit PDF
- SSO & on-prem