Basket
One therapy across several indications, each cohort with its own enrolment and stopping rules.
ClinPrecision, by Novatra AI
Built for the trial you actually designed.
Basket, umbrella, platform or a single-arm Phase II. ClinPrecision runs data capture, operations, documents and quality on one auditable record, with AI that proposes and people who approve.
EDC · CTMS · eTMF · QMS on one platformDesigned for 21 CFR Part 11CDASH / SDTM mappingMedDRA & WHODrug coding
Platform
Most trials run on fifteen systems. Yours can run on one record.
The usual stack sells integrations between products. When one architecture owns data capture, operations, documents and quality, the hand-offs disappear and behaviour follows the data.
Industry default
Many systems, reconciled by people
- Every interface is a project with its own validation
- Teams reconcile exports instead of reviewing data
- A change in one system reaches the others late, or never
ClinPrecision
One event-sourced record, every module reads it
- Data capture, operations, documents and quality share one audit trail
- Cross-module rules fire on events, not on nightly jobs
- Inspection readiness comes from the architecture
A deviation becomes a quality event before anyone writes an email.
- 09:14:02.118Data capture
ProtocolDeviationRecordedA visit falls outside its window at Site 104.
- 09:14:02.164Quality
QualityEventOpenedOpened with the subject, site and visit already attached.
- 09:14:02.201Quality
CapaInitiatedRouted to the site's quality owner, with a due date.
- 09:14:02.247Training
TrainingAssignedRetraining on the visit schedule goes to the affected roles.
4 events · 129 mseach signed to the user who caused itappend-only, never edited in placeexample sequence
Master protocols
Every study is a set of cohorts. A standard trial is just a set of one.
In about a decade, master protocols went from a niche design to a large share of oncology trials. ClinPrecision models cohorts from the data model up, so a platform trial uses the same screens as a Phase III, with no special mode and no add-on.
Umbrella
One disease, a shared biomarker panel, and patients routed to the matching cohort.
Platform
Arms added and dropped mid-study at interim looks, under governed design changes.
SMART
Patients re-assigned at decision-point visits based on how they responded.
Works with your IRT. Randomisation and drug supply stay in the IRT you already trust. ClinPrecision consumes its assignments and keeps blinded data blinded.
How the AI works
The system proposes. A person approves.
During study build, the platform suggests eligibility criteria, CDISC mappings and cross-form checks. Every suggestion carries a score you can take apart, and the score decides who has to look at it.
AUTO_APPLIED
Score of 0.92 or higher. Applied, logged, and reversible.
PENDING_REVIEW
Below 0.92. Waits for a named reviewer to accept or reject it.
UNMAPPED
Nothing to bind to. A person maps it by hand.
"Date of onset"→AE.AESTDTC
0.951AUTO_APPLIED
context ×0.25
semantic ×0.20
assessability ×0.20
dependencies ×0.15
history ×0.10
catalogue ×0.10
score = Σ weight × part · example values
By job
What changes for the people running the trial.
assemble the database
Study build
Review a proposed database, eCRFs and edit checks derived from the protocol, then approve.
eyeball every discrepancy
Data management
Validation runs continuously. People judge the residuals that need judgement.
travel and spreadsheets
Monitoring
Sites ranked by risk, so the CRA knows where to look first and why.
"where do I file this?"
Documents & quality
Filed on arrival, with the gap, the deviation and the suggested fix in one place.
Why now
Three conditions arrived at the same time.
Models can be trusted to propose
Clinical reasoning and protocol reading are now good enough to draft, as long as a person reviews.
Regulators are converging
FDA, EMA and ICH guidance accommodates AI-supported work when the audit trail and explanation come first.
Sponsors are out of patience
Faster start-up, cheaper monitoring and fewer queries are the levers they need, and the old stack can't pull them.
Founder
“I've seen how this stack is built from the inside, and why it can't carry the next decade of trial designs.”
Narendra Nath SarkarFounder & Chief Architect, Novatra AI · 20+ years of enterprise architecture in pharma, financial services and healthcare
ParexelMedidataOraclePerceptive
Bring a protocol. We'll show it running.
A working session on your own design: cohorts, visits, forms and the checks the platform would propose for it. We reply within 1–2 business days.
Phone · Mon–Fri, 9 AM–6 PM CT1-469-238-1258
Designed for 21 CFR Part 11ICH E6(R3)GxP-ready architecture