What is AccuraTrials?

AccuraTrials is the intelligence layer for clinical trials: a semantic and ontology model of the study, built on CDISC USDM and OpenStudyBuilder, that a protocol is read into. From that model the platform drafts eCRFs, compiles plain-language rules into tested edit checks and derives the visit schedule, with every output citing the protocol clause behind it. Electronic data capture built to 21 CFR Part 11 runs on top of the same model.

How does AccuraTrials reduce build times?

AccuraTrials reduces build times by generating eCRF drafts directly from the protocol model rather than from a vendor ticket, and by compiling plain-English validation logic into versioned deterministic rules, all of it for human review before activation.

How does AccuraTrials support inspection readiness?

The evidence is written while the work happens rather than assembled afterwards: edit checks catch anomalies before a form is saved, every change carries its reason, and AI-assisted audit review surfaces patterns in audit activity, such as missing change reasons or repeated late edits, while there is still time to act on them.

Is AccuraTrials 21 CFR Part 11 compliant?

Part 11 compliance is something a sponsor demonstrates for its own system rather than a badge a vendor holds, so the honest answer is that AccuraTrials is built to support it from the ground up. Every electronic signature requires credential re-authentication, and every data change writes to a hash-chained, tamper-evident audit trail with timestamp, user, prior value, new value, and reason.

How long does study setup take with AccuraTrials?

AccuraTrials accelerates study setup. The AI protocol parser generates eCRF drafts from a protocol PDF in minutes for your review, and the 8-tab study wizard handles sites, visits, randomization, and permissions in a single session. Launch timing depends on your review and validation process.

Who builds and amends the study, the sponsor or a vendor?

The study team does. The AccuraTrials Form Builder, Validation Rule Builder and Branching Logic editor are visual interfaces with no XML and no scripting in them, and the AI Rule Suggester proposes edit-check logic from a constraint described in plain English for a person to approve. Amendments are deployed by the team running the study rather than through a vendor change order.

Who is AccuraTrials for?

AccuraTrials is built for sponsors running a program of studies, contract research organizations (CROs) running studies for many sponsors, and biotechs running a first trial, across Phase I to Phase IV and observational designs.

What export formats does AccuraTrials support?

AccuraTrials supports PDF, Microsoft Excel, CSV, ODM XML (CDISC standard), and tab-delimited export. AccuraTrials includes a dedicated regulatory export endpoint for FDA and EMA submission packages with complete audit trails.

How does randomization work in AccuraTrials?

AccuraTrials supports stratified randomization with configurable stratification factors and blinded study arms, with an unblinding workflow for when a treatment assignment must be revealed. AccuraTrials RTSM/IRT kit and shipment tracking is integrated.

Where is AccuraTrials based?

AccuraTrials is based in Madison, Wisconsin, USA. It was founded in 2025 by co-founders Bhuwan Sai Kosuru (CEO) and James Gui (CTO) to build the semantic and ontology layer that lets AI work on clinical trials traceably enough for a regulator to review.

AboutMadison, 2025

The system of action.

A semantic layer that lets AI work on a study without inventing anything, and the platform that runs on top of it. Two founders, one live paid trial, and a thesis about which part of a clinical trial is still done by hand.

Talk to us How the model is built

Thesis03 eras

Two companies made the record digital. Neither of them runs a trial.

A system of record holds what happened, and both of the last two generations of clinical software are very good at it. What a trial still needs done by hand is everything in between.

Paper to EDCMedidata digitized the dataCase report forms stopped being paper and the record became electronic. The company was acquired for $5.8B. What changed was where the data lived.

On-prem to cloudVeeva moved the documentsThe regulatory binder, the site file, the submission: hosted, versioned, shared across a study team. A company worth north of $30B today. What changed was where the documents lived.

Manual to AIAccuraTrials runs the trialBoth of those hold the record faithfully. The work of a trial — reading the protocol into every system, catching the deviation, pricing the amendment, proving all of it afterwards — is still done by people, by hand, once per system. That work is what we are building software for.

What one model of a study changes

Costwhat it takes now

The tax on a document no machine can read.

The protocol is the authority for everything that happens in a study, and it arrives as prose. Every system downstream of it is configured by a person re-typing that prose, and every re-typing is a place two systems can come to disagree about the same subject.

25–30%of a trial’s cost is people checking other people’s work

$535Kto implement one substantial protocol amendment

3.5substantial amendments per Phase III protocol, on average

None of that is waste anybody chose. It is the price of one document being re-interpreted by hand into every team, vendor and system that needs it — and it is paid again on every amendment.

~$500Kin lost prescription sales for every day a trial slips

~$15Mgone per month of delay, plus $1.6M+ spent keeping the study crawling

15–18months recoverable across a program when every phase re-uses one model

The first two figures are what the industry already pays. The third is arithmetic on re-use rather than a result we have measured — we have one live study, and we would rather say that here than be asked on the call.

Amendment cost and frequency: Tufts Center for the Study of Drug Development. Trial-delay figures are widely cited industry estimates. Neither set is a measurement of this platform.

Founderstwo of us

Two founders, one thesis.

One of us has spent years making documents machine-readable. The other ships software into environments where a validation gap is a finding.

Bhuwan Sai Kosuru

Co-founder and CEO

Built an AI data company from zero to $500K in revenue in under a year, solo, with four YC-backed startups as customers.

Bhuwan has spent his career on the hard part of applied AI: getting messy real-world documents into a form a machine can act on without making things up. He owns the clinical and regulatory side here, which means sitting with data managers and monitors until the system does what they need rather than what demos well.

Previously at Giga (YC S23) and atlas.so (YC W22). IIT Kharagpur ’24bhuwan@accuratrials.com

James Gui

Co-founder and CTO

Built clinical-trial infrastructure end to end inside regulated health-data and diagnostics environments, then built the AccuraTrials platform and EDC now running in a live study.

James has shipped software into places where a validation gap is a finding and not a bug ticket: health-data systems at Extract Systems, cancer diagnostics at Exact Sciences. He owns the platform here: the protocol model, the EDC, the audit architecture, and the rule that nothing reaches a study without a human approving it and a source line behind it.

Previously: health-data systems at Extract Systems, cancer diagnostics at Exact Sciences. jgui@accuratrials.com

Edgewhy this holds

Three things that compound.

ProvenanceProvenance is the architecture, not a featureIt cannot be switched off to go faster, because there is no path through the system that skips it. Every artifact carries the clause behind it, so trust compounds study over study instead of being re-established from scratch each time.

Re-useStudy logic is reusable, so each study is cheaper than the lastA protocol model is not consumed at go-live. The next phase of the same molecule starts from it, which means the modelling cost is paid roughly once across six to eight phases rather than once per study.

ChannelCROs are the channel, not the competitionCoordination at the scale of dozens of people per study is how the industry’s hands actually work today. We are not trying to replace them — we are building the system they run on.

Statustoday

Where we actually are.

In a study: Live in a paid Phase I trial today. Accura Studio built it and the AccuraTrials EDC is running it.
On the base: Accura Cortex is USDM-native and built on OpenStudyBuilder. Accura OS reaches the EDC today, with four more systems connected as prototypes.
Where: Madison, Wisconsin. Founded 2025. Two founders, no outsourced engineering.

Purposewhat it is for

What a faster trial is actually for.

~7,000rare diseases described

~95%of them with no approved treatment

~300Mpeople living with one

Before the Orphan Drug Act in 1983, fewer than forty orphan drugs existed. Since then, more than six hundred. The science did not change that year — the economics did. What a trial costs decides what a company is willing to attempt, and in North America alone more than 250,000 life-years are lost every year waiting on approval.

More cures, sooner. That is the entire reason to make this faster.

Nextthe ask

Come look at the model.

Thirty minutes with both founders. Bring a protocol if you have one.

Book a walkthrough See the platform