TrialMind: an agentic AI platform that runs the full clinical-trial lifecycle — literature synthesis, protocol design, eligibility criteria and cohort extraction, data-science co-pilot for biostatisticians, and regulatory drafting Web. Spun out of the University of Illinois ML-for-Healthcare lab of Jimeng Sun, ex-Global Head of AI Research at IQVIA Web.
Keiji is a rare revenue-funded generative-AI company in clinical trials — on pace for $5–7M ARR this year with paying pharma and biotech customers (Regeneron, Structure Therapeutics, Loomis, Garden Health, AbbVie) and no institutional capital raised — sitting on top of a proprietary trial-data pipeline (850K protocols from 14 countries linked to 1.3M publications) built inside a top academic AI-for-health lab. The wedge is bespoke clinical-data access plus a domain-tuned agent stack rather than a generic LLM wrapper: contracts run $200K–$1M, expansion is happening inside customers (Garden went from 1 to 3–4 SOWs), and the platform ships an MCP server so it can co-exist with corporate Claude Code deployments Internal. The drag is founder posture — Jimeng, the CS-professor CEO, is not urgently seeking capital and has been slow to engage on fundraising process; StoryHouse’s posture is to keep the option warm at $800K–$1M in a $2–3M-sized syndicate rather than push for a full round.
Clinical-trial gen-AI is one of the fastest-growing enterprise AI verticals: 25–30% CAGR across three independent research houses, anchored by the R&D-cost pressure inside pharma (average trial costs $50M+ per Phase 3) and the multi-year push toward decentralized and adaptive trials. Keiji is entering while enterprise buyers are still calibrating agentic-AI budgets, and the moat that matters — access to structured trial protocols and outcome data — is being built in a category where scraped-text LLMs have limited leverage.
| Player | Positioning | Funding / Stage | Edge vs. them |
|---|---|---|---|
| Keiji AI (TrialMind) | Agentic platform across the full trial lifecycle; proprietary 850K-protocol + 1.3M-publication data pipeline; MCP-first architecture Internal | Bootstrapped · $2M ARR → $5–7M target | — |
| Tempus AI | Public precision-medicine / trial-matching platform; acquired Deep 6 AI to add 30M patients across 750+ sites Web | Public (NASDAQ: TEM) | Tempus is oncology-first with clinico-genomic data; Keiji targets the operational trial layer regardless of TA |
| Unlearn.ai | Digital twins for control-arm augmentation in trials Web | Series C, $15M+ ARR range | Different wedge: Unlearn is a statistical trial-design product; Keiji spans the full lifecycle including exec/data-science co-pilot |
| Owkin | Multimodal patient-data federation + AI biotech; Paris/NY, ~$300M raised Web | Series B+ | Owkin runs its own biotech pipeline and hospital federation; Keiji is a pure enterprise SaaS/agent play for sponsors and CROs |
| Anthropic Claude / OpenAI (in-house) | Enterprise LLMs used directly by pharma dev teams (“Claude Code”) Internal | N/A (platform) | Real threat — but Keiji ships an MCP server so it composes with, rather than competes head-on against, corporate Claude deployments |
| Trials.ai | Named direct competitor per Tracxn; earlier-stage Web | Seed | Narrower feature scope; less academic-lab data-moat depth |
Moat: a proprietary trial-data pipeline (850K protocols, 14 countries, linked to 1.3M publications, structured to ICD/MeSH) built inside the UIUC lab plus a domain-specific agent stack. The MCP-server strategy positions Keiji as a data-and-workflow layer that composes with enterprise LLMs rather than competes with them — a defensible posture against Claude/OpenAI DIY.
The category is consolidating around strategic buyers: Thermo Fisher closed the $8.875B Clario acquisition in March 2026, IQVIA acquired AI analytics firm Whiz, and Veeva acquired the Ostro patient-engagement platform for $100M in March 2026 Web. Tempus AI absorbed clinical-trial-matching Deep 6 AI to bolt patients-and-sites onto its precision-medicine offering. The plausible acquirors for Keiji are IQVIA (natural Jimeng-alumni buyer), Veeva, Medidata/Dassault, Tempus, Thermo Fisher, and Icon; a strategic pharma buy (Regeneron, AbbVie) is possible but less common in this workflow-tooling category. A revenue-multiple exit at 8–12x forward ARR is the base case if ARR compounds through $25–50M, and the return math works cleanly from a sub-$40M entry.
Brandon apologized for a delayed follow-up and sent the deck plus demo access via keiji.ai. Confirmed Keiji is not attending BIO San Diego and explicitly deferred all fundraising discussion to Jimeng — a soft signal that Brandon is not the decision-maker on capital.
Keiji is transitioning from co-development pilots to renewable annual licensing. Data Science Platform (main product) has 100–150 active users, Garden renewed 50 seats at $7,500 each, Structure Therapeutics and Loomis at 10 seats each on-prem. Serious fundraising exploration but no urgent timeline. StoryHouse indicated a possible $800K–$1M check, willing to co-lead a $2–3M round; would not lead a $5M+ raise alone.
Reconnect after several months of limited engagement. Confirmed $5–7M ARR target unchanged; three products: data-science platform (main), literature review, eligibility criteria generation. Verdict: interesting business, but CEO is not engaged on capital and not worth pushing hard.