Sarivera

Big-pharma clinical execution, for sponsors that can't hire it.

We run early-phase drug trials for small sponsors with the intelligence of a large sponsor: design, endpoints, analysis, regulatory strategy, execution.

The problem

Small sponsors develop comparable molecules with less success.

Investigational drugs reaching approval: 28 percent for small-company sponsors, 42 percent for large-company sponsors, after adjustment for therapeutic area, molecule type and pathway novelty.

A large sponsor incumbent hires hundreds of millions of dollars of clinical expertise to succeed. A small emerging biotech or academic sponsor can only afford millions.

What we do

We run the drug trial, on systems we build and own.

Software vendors and conventional CROs each solve half of it. SaaS vendors build AI systems but never run a trial. Conventional CROs run trials but in-license their AI systems, and their expertise stays siloed in the heads of staff who often move on. Neither builds intelligence that accumulates.

SaaS vendors

License software to the organisations that run trials, but never operate a study themselves. The AI never learns from a study being run.

Conventional CROs

Execute the trial, but license their AI systems from a vendor. People learn from each study, and the institutional knowledge leaves when they do.

Sarivera

Operates the study on AI systems it builds and owns. We build durable intelligence — the processes, knowledge and decision support currently siloed inside big pharma.

How a trial runs

One governed system, from protocol to readout.

Protocol
Study build
Activation
Conduct
Amendment
Readout

Sarivera

One AI system. We build it, we run it.

Protocol ingested read, not retyped

Generated from the protocol

  • Schedule of assessments
  • eCRF and EDC schema
  • Edit checks, visit windows
  • Endpoints · TMF · safety
  • Site records
  • Grants and contracts
  • Budget
Operations, running together
  • Data capture and edit checks
  • Queries raised and resolved
  • Safety events and reporting
  • Monitoring record
  • Site payments on milestones
  • Quality findings
One state — not six, reconciled later
Artifacts regenerated
Data captured under v5 is still read as v5
Analysis on governed data bound to a snapshot
Every one of these writes to the same versioned record immutable snapshots · sha256 · append-only event log · actor, timestamp and lineage on every change

Sites

  • Investigator
  • Study coordinator
Activated and trained
Enters visit and module data Answers queries Investigator signs off Validated before it commits
An amendment reruns the build — forms, checks, visit windows and site budgets update together

The platform

Regulator-friendly by design, not by retrofit.

The EDC run screen: subject enrolment, eligibility module, and the governed operational state alongside.
The platform conducts the study. Subject enrolment, visit and module data entry, validation before anything commits. The EDC schema is generated from the protocol, not built by hand afterwards.
Analysis Studio: saved analyses on the left, an evidence-bounded response on the right.
Analysis runs on governed data, during conduct. A question in plain language becomes a governed, re-runnable analysis bound to a versioned snapshot — with the judgment calls surfaced to you rather than absorbed into code.
A synthesised view of the complete study state: status, protocol and amendment state, governance, sites, EDC and data.
The whole study, in one answer. Protocol state, amendments, governance, sites, safety, data quality and financials — synthesised across functional boundaries, because they were never siloed in the first place.

Immutable artifact lineage

Every governed data artifact is versioned, hashed, and traceable to its source and its full history.

Deterministic guardrails

Agents propose. Deterministic executors apply changes under human guardrails. The AI model never gets the last word.

Epoch-correct history

Every operation is read against the protocol version in force at the time. History can be reconstructed exactly.

Still to build

Controlled released EDC versions with effective dates and retirement, governed data artifacts, an amendment ledger and baseline freeze at study initiation are in place. The remaining 21 CFR Part 11 and ICH E6 controls are not: EDC-entry audit trail, electronic signatures, controlled document states, and upload metadata.

The founder

I use technology to level the playing field.

David Johnson

David Johnson, Ph.D.

Founder and Chief Executive Officer

I see problems in biotech and know how to fix them. I have delivered difficult technologies with teams I hired — papers in Science and Nature Biotechnology, inventions worth billions of dollars.

I have experienced firsthand how siloed clinical expertise impedes emerging therapeutic technologies.

I see AI as an opportunity to fundamentally change the way clinical trials are run, and finally level the playing field with big pharma.

Natera
Founding COO. Natera built the first and leading non-invasive molecular diagnostics for reproductive health and oncology. Now public, with over $1bn in diagnostic product revenue.
GigaGen
Founder and CEO. I invented recombinant polyclonal antibodies — thousands of distinct antibodies manufactured as a single drug using microfluidics. Grifols acquired the company in 2021. Three drugs from that work are in active clinical studies today.
GigaMune
Co-founder. I invented a bioinformatic method for identifying hundreds of novel in vivo CAR-T delivery vectors for oncology. Several candidates are in preclinical development, supported by a $2.4M Fast Track SBIR from the NCI.

Bring us a protocol.

The entry point is a request for proposal on a specific early-phase study. We return a demonstrated study build against your draft protocol in days rather than the weeks a conventional CRO takes to return a bid.

[email protected]