Your data is
the asset.
Evidia turns the real-world data you already own into governed, audit-ready evidence — without handing it to a vendor's dataset.
Now accepting design partners — emerging and mid-size biotech, rare disease and specialty first.
YOUR TENANT
External control-arm dossier
study: RD-201 · rare disease · ontology v0.2.0
Your claims + EHR extracts mapped to OMOP, bronze → gold.
Purpose, de-identification and ontology checks enforced.
Agent assembles cohort, control arm and analysis — every claim cited to its source.
Your team signs; the package is fingerprinted into the vault.
You already own the data. It's just not usable yet.
Most biotechs sit on years of claims, EHR extracts, registry and trial data. Turning any of it into evidence a regulator or payer will accept is where programs stall.
Incompatible formats
Every source speaks a different dialect — different codes, different visit logic, different missing-data habits. Before any analysis, someone has to reconcile all of it by hand.
Privacy risk
Patient-level data can't just be pooled, emailed, or uploaded to a vendor. De-identification has to be methodical, documented, and defensible — or the data never leaves the building.
Scrutiny it won't survive
Regulators and payers ask where every number came from. Analyses assembled in notebooks and spreadsheets can't answer that — so evidence gets discounted, or redone at services rates.
The legacy answer is to rent answers from someone else's dataset — a marketplace built on their patients, their definitions, their roadmap. Evidia starts from the opposite premise: harmonize and govern the data you already own, and the evidence is yours.
One foundry. Six products.
One governed foundation: your data, harmonized once, with an ontology enforced on every query, agent run, and dossier. Explore below.
Evidence Foundry
The platform everything else stands on.
Connect the data you already hold — claims, EHR, registries, Snowflake, Databricks, Foundry, or files — and get it back harmonized to OMOP, de-identified, and quality-scored, inside your own isolated tenant.
- Self-service connectors — tested against your environment before anything moves.
- OMOP harmonization — bronze → silver → gold pipelines turn incompatible extracts into one queryable model.
- De-identification + quality — HIPAA Safe Harbor, tokenization, and visible completeness and lineage scores on every table.
ControlArm Studio
Our flagship: external control arms, productized.
For rare-disease and specialty trials where a randomized control arm is impractical or unethical, ControlArm Studio builds the external control arm from your own harmonized data — delivered as one signed, fingerprinted evidence package.
- Cohort to control arm — protocol-aligned eligibility, index-date logic, and balancing, defined once and versioned.
- Analysis included — estimand-aligned analysis runs on the governed layer, not a spreadsheet export.
- Priced per study — $150–300k, budgeted like a line item, not a services negotiation.
Agent Suite
Governed AI agents that do the drafting — you keep the pen.
Purpose-built agents handle the repetitive core of evidence work — study design, cohort QA, dossier drafting, feasibility. They reason over the governed ontology, never free-form over raw data, and nothing ships without your sign-off.
- Four working agents — study design, cohort QA, dossier drafting, feasibility, each scoped to governed tools and data slices.
- Ontology-bound and cited — every result names the ontology version and sources behind it. No unattributed numbers.
- Human sign-off, always — agents draft; your scientists review and release. Fail-closed: an unvalidatable run stops rather than guesses.
Scout
Ask your own data questions. Get governed answers.
Always-on evidence intelligence over your harmonized data. Ask feasibility and landscape questions in plain language — get answers computed against the governed layer, with the cohort logic shown.
- Plain-language questions — a governed query, not a ticket to the data team.
- Feasibility in minutes — count patients, sites, and event rates before you commit to a protocol.
- Transparent by default — every answer shows its cohort definition, data vintage, and ontology version.
Ontology Forge
The governed knowledge layer under everything.
The contract that keeps every concept consistent across sources, studies, and agents: OMOP concepts, estimand frameworks, and regulatory definitions — versioned and enforced.
- 40-class governed ontology — clinical, estimand, and regulatory concepts as one versioned model (v0.2.0).
- Enforced, not documented — agents and dossiers run against the ontology; definitions can't drift.
- Version-pinned and extensible — every output records its ontology version for reproducibility; extend it with your therapeutic-area concepts.
Part 11 Vault
Evidence that can survive an audit — because it was born in one.
The regulated-evidence layer: e-signatures, an immutable audit trail, and validation documentation generated as work happens — not reconstructed before an inspection.
- E-signatures — named sign-off on cohorts, analyses, and packages, bound to the exact artifact signed.
- Immutable audit trail — who did what, when, under which ontology version and data vintage, in a tamper-evident chain.
- Fingerprinted and validated — cryptographic fingerprints you can re-verify anytime, with validation documentation alongside.
See it in action
Two miniature walkthroughs of the working product. Both run on illustrative sample data — click through them.
Scout: ask your data
Pick a sample question. Scout answers against the governed layer — with the ontology version and sources shown.
1,248
patients match · sample dataset · vintage 2026-Q3
CITATIONS
- Treatment-naïve: no systemic therapy in 365d lookback (RxNorm)
- Age ≥ 65 at index · ≥ 2 prior lines of therapy
EVIDENCE
- gold.cohort_rd201_v3 · 1,248 rows
- query log q-8841 · reproducible
Illustrative sample output. Real answers compute against your tenant's harmonized data.
18.4%
12-month event rate · n = 1,248 · sample dataset
CITATIONS
- Event: protocol-defined composite endpoint (estimand EF-02)
- Censoring per protocol §4.3 · Kaplan–Meier
EVIDENCE
- gold.analysis_rd201_km · run a-2210
- query log q-8842 · reproducible
Illustrative sample output. Real answers compute against your tenant's harmonized data.
3 sites
above 12 patients/month · sample dataset · 2026-Q3
CITATIONS
- Enrollment rate: consented / site / month, de-identified
- Site identities masked in sample output
EVIDENCE
- gold.site_enrollment_q3 · 42 sites
- query log q-8843 · reproducible
Illustrative sample output. Real answers compute against your tenant's harmonized data.
ControlArm Studio: dossier walkthrough
Three steps from cohort to a signed, fingerprinted dossier — the productized per-study delivery.
Build cohort
Protocol-aligned eligibility is defined once and versioned — index-date logic, washouts, and exclusions pinned to ontology v0.2.0.
- Eligibility compiled from protocol §3 → 1,248 patients
- Balance diagnostics pass on 14 covariates
Estimate control arm
Estimand-aligned analysis runs against the governed layer — not a spreadsheet export — with every number traceable to its source.
- Doubly-robust estimate: 18.4% event rate (12 mo)
- Sensitivity analyses included in package
Fingerprinted dossier
One signed artifact: cohort, methods, results, provenance, and e-signatures. Verify the fingerprint anytime.
study: RD-201 · rare disease ontology: v0.2.0 cohort: n=1,248 · def v3 estimand: EF-02 · 12-mo event rate result: 18.4% (doubly-robust) fingerprint: 9f2c…a41d (illustrative) signatures: sponsor · biostatistician
Illustrative sample manifest — not a real study or fingerprint.
Start with one study. Keep the platform.
Deliberately narrow: get your first signed evidence package out the door, then decide how far the platform goes. No open-ended consulting.
Design-partner pilot
One dataset, one study, one signed evidence package — judge us on evidence, not slides.
Federated benchmarking
Compare cohorts against cross-company aggregates without pooling raw data — only k-anonymous aggregates move.
Tenant onboarding
Your isolated environment — your tenant, your keys, your deployment model — connected with you, so your team runs the platform.
From raw extracts to signed evidence in five steps
Each step produces something concrete your team keeps — including if you stop after the pilot. Select a step to see what it does.
Step 1 of 5Connect
Your Snowflake, Databricks, Foundry, or file sources are linked and access-tested inside your tenant.
You get: an inventoried, working data estate.
The Foundry pipeline, visualized
Watch a source travel through harmonization, de-identification, and governance into signed evidence.
Nodes light in sequence as data moves down the pipeline. Every stage is audit-logged against ontology v0.2.0.
Deployment your way
The same platform, three ways to run it. Choose by your data-governance posture — your data never has to move to us.
Pooled SaaS
Fastest start. Your tenant runs in our managed environment with hard tenant isolation and your own encryption keys — provisioned in days, not quarters.
Private single-tenant
A dedicated silo: your own environment, network boundaries, and key material — for sponsors whose data policies rule out shared infrastructure.
Inside your AWS account
The platform deploys into your own cloud, connected to your data where it already lives. Maximum control; your security team reviews our footprint in your account, not a brochure.
Isolation is architectural, not contractual: per-tenant encryption keys, tenant-scoped data paths, and a control plane your admins can audit — whichever model you choose.
Why teams choose Evidia
Dataset marketplaces sell you their patients. Legacy RWE vendors sell you their consultants. We built the third option.
Your data is the product's object
Everything in Evidia works on the data you already own — harmonized once, then reused across every study. We bring no competing dataset and no incentive to steer you toward one.
Self-serve, not services-led
Provision your tenant, connect your sources, run your studies from the control plane. Our team helps you start; the platform is how you scale — no standing army of consultants required.
Pricing you can put in a budget
Per-study dossiers at a published $150–300k anchor and pilot pricing on this page. You can take Evidia to a budget committee without a six-week procurement excavation.
Every output is auditable
Ontology version, data vintage, provenance, and a fingerprint on every package — plus human sign-off before release. When a regulator or payer asks "show me," you can.
Published anchors. No hidden maze.
Three numbers cover most questions. Final scope is confirmed in writing during scoping — before anyone signs.
The front door: one dataset, one study, one signed evidence package.
- One source connected & harmonized
- One signed, fingerprinted dossier
- Your harmonized data stays yours
External control-arm dossiers as a product after the pilot.
- Cohort, control arm, and analysis included
- Fingerprinted, re-verifiable package
- Priced before work starts
The full foundry — priced to your deployment model and tenant footprint.
- All six products, one governed platform
- Pooled, private silo, or your AWS account
- Admin-controlled services, per tenant
Anchors are planning numbers, confirmed in writing during scoping. If your study doesn't fit them, we'll say so before you commit.
Questions sponsors ask us first
Yes — fully. You hold your tenant's encryption keys, control access through purpose-based policies, and can export your harmonized datasets and evidence packages at any time, including at pilot end. We never pool your raw data with another customer's; federated benchmarking moves only k-anonymous aggregates.
You choose: pooled SaaS with hard tenant isolation, a private single-tenant silo, or inside your own AWS account. In the third model your data never moves to us — the platform runs next to it, under your cloud governance.
It's a pipeline stage: HIPAA Safe Harbor identifier removal, date shifting and generalization where required, and tokenization so longitudinal linkage survives without exposing identity. Every release carries a residual-risk report your privacy team reviews first.
One signed, fingerprinted artifact: cohort and control-arm definitions, estimand and methods, results, data vintage, the exact ontology version, full provenance for every number, and the e-signature record. The fingerprint can be re-verified later to confirm the package is byte-for-byte what was signed.
Straight answers: the Part 11 Vault provides e-signatures, an immutable audit trail, and validation documentation designed against 21 CFR Part 11. HIPAA-aligned design, with a BAA available in pilot contracting. SOC 2 is on our roadmap — we won't claim a certification we don't hold, and we'll show you where the program stands during diligence.
Scoped to first evidence in weeks, not quarters: connect and harmonize one source, run one study through ControlArm Studio, sign one package with your team. The main variable is your data's readiness — which is why the pilot starts with a source assessment.
Snowflake, Databricks, and Palantir Foundry, plus file-based sources — claims extracts, EHR exports, registry feeds — in common formats. Connectors are self-service and tested against your environment before anything moves. Already in OMOP? Harmonization starts from silver.
You keep everything the pilot produced: harmonized datasets, the signed evidence package, and the definitions behind it. Most teams convert to the annual platform or commission per-study dossiers — but the pilot stands on its own if you don't. Nothing is held hostage to a renewal.
Build your design-partner brief
Three quick steps, no email gate. Compose a pilot brief you can copy, download, and take to your team.
Your pilot brief
Take it to your team — or send it to us and we'll come prepared.
Walk through the three steps and your design-partner pilot brief will appear here, ready to copy, download, or send.
Talk to us
Pilots, live demos, partnerships — a human replies, not a drip campaign.
Tell us what you're working on and where your data lives today. We'll come to the first conversation already thinking about your study design.