For pharma innovation and digital R&D teams
Build clinical AI with Keiji AI
We help pharma teams build their own clinical AI, from fine-tuned models to agents, on the systems they already use.
Our tutorial explains why the right training data and a good evaluation set matter more than model size. Read the tutorial
What the engagement includes
What innovation teams get when we build with them.
Clinical domain expertise
General AI tooling is quick to set up. Knowing what a validated clinical data standards (CDISC) workflow requires, and when AI output can no longer be defended, takes years of experience.
Existing agent infrastructure
Our agents and tool infrastructure are already built, so the engagement starts from working clinical tools rather than from scratch.
Capability your team owns
We support your team's mandate to build. The researchers who wrote the papers do the work themselves.
What each one includes
- Validated workflows for CDISC tabulation (SDTM) and analysis (ADaM) datasets
- Statistical analysis plans (SAPs) broken into agent tasks a biostatistician recognizes
- Clear limits on where AI output needs human review
- Regulatory constraints designed in rather than discovered late
- Production agent infrastructure and a library of clinical Model Context Protocol (MCP) tools
- Built on whatever platform you have already chosen
- Custom skills and validated pipelines on your infrastructure
- A working system in weeks
- Built on your infrastructure and owned by your team
- Knowledge transfer is part of the deliverable
- A defined scope with a delivery date
- The researchers who published the work do the work
How to evaluate us
Innovation teams usually evaluate partners on published research and working systems.
Published research
Peer-reviewed work on large language models (LLMs) and clinical trials that you can read before engaging us.
Working systems
We show running systems and published papers, not slide decks, and we do not require a software license.
Your platform
We build the domain layer on the platform you already selected, rather than requiring you to adopt ours.
Technology partners
Cloud, compute, and open-source platforms TrialMind is built on.
Engagement details
- Engagement model
- A scoped engagement, not a license. TrialMind may not appear in the statement of work
- Deliverable
- Custom agents and validated pipelines on your infrastructure, plus knowledge transfer
- Capabilities
- Large language model (LLM) fine-tuning, clinical agent development, clinical Model Context Protocol (MCP) tool library
- Platforms
- The platforms you have standardized on, including major cloud and data platforms and open-source scientific tools
- Who does the work
- The researchers who published the papers, with no separate delivery team
- Ownership
- Your team owns what is built; rights to generalizable components are agreed up front
Frequently asked questions
Do we have to buy TrialMind?
No. This is an engagement, not a license. The deliverable is custom agents and validated pipelines built on your infrastructure, and TrialMind may not appear in the statement of work. If a licensed product is the better answer for a given workflow, we will say so.
Who actually does the work?
The researchers. There is no handoff from the people who wrote the papers to a separate delivery team, which is usually where domain understanding gets lost.
How are engagements scoped?
Every engagement has a defined scope and a delivery date. We do not take open-ended staffing arrangements, because they rarely produce something your team owns.
What happens to what we build together?
Your team owns the capability built on your infrastructure. We retain rights to generalizable components and methods, and we agree that boundary in writing at the start.
How is this different from a consultancy?
We are a small team of researchers and engineers who have already built production agent infrastructure for clinical development, and we are accountable for whether the system works. You can check our publication record before signing anything.
Which platforms do you build on?
Whatever your organization has standardized on. In practice that has meant the major cloud and data platforms alongside open-source scientific tools. We work inside your architecture.
Discuss your project
Talk with the researchers who would do the work.