Why this program exists
Many promising laboratory signals in oncology never reach patients because the Phase I/II gap between bench and bedside remains expensive, slow, and unevenly navigated. Small academic groups with strong translational science often lack the trial-design infrastructure that large sponsors take for granted, while the methodologies used to design those trials are usually published behind paywalls and rarely subjected to community benchmarking. We believe this is exactly the kind of methods-and-open-benchmarks work a public-benefit institute should take on.
Our approach
Our research applies modern AI methods — large-language-model pipelines, structured extraction, and computational biomarker analysis — to the specific questions investigators face when designing first-in-human oncology trials: which investigators and sites are best suited, how to allocate starting-dose cohorts, which biomarker strata are most likely to concentrate signal. We publish our methods, our benchmarks, and our negative results, so that other academic groups can build on them without having to pay for access.
What we ship
Our current focus is a methodological manuscript on AI-assisted Phase I/II trial design. The outline for this paper is published on this page as a public roadmap so collaborators and reviewers can engage before submission. As methods mature we release them openly — first as reference code, later as benchmarked datasets for community comparison. We do not accept paid sponsor engagements from commercial pharmaceutical or biotechnology entities; every collaboration in this program stays academic and non-exclusive.
Who benefits
The immediate beneficiaries are the academic oncology community and clinical-trial investigators who need rigorous but accessible methodology. The ultimate beneficiary is the patient who, a few years downstream, reaches a well-designed first-in-human trial of a therapy that otherwise would have stalled. Our outputs are global goods: we serve the field, not a single sponsor.
How to engage
We welcome academic collaborators, review feedback on our methods papers, and correspondence from groups exploring similar questions. Reach out to introduce a collaboration, share a manuscript for discussion, or ask about joining a working group.
Outputs
Shipped(1)
Released under open license.
FIH Scout ↗
Oncology first-in-human trial intelligence — a live web platform exploring AI-assisted investigator and site ranking for early-phase oncology studies. Open for academic collaborators to try.
Syllabi & outlines(1)
Public roadmaps — published before full content so collaborators can shape them.
FIH Trial-Design Methods Outline
A public roadmap for the methodological manuscript on AI-assisted Phase I/II oncology trial design. We publish the outline before the paper so collaborators and reviewers can engage with the methodology early.
- Problem framing — where current Phase I/II trial design decisions lose efficiency
- AI-assisted investigator and site selection methodology
- Starting-dose cohort allocation under sparse prior information
- Biomarker strata identification and validation approach
- Open-benchmark protocol — data, code, and negative-result release policy
Work with us
We are always open to new collaborators and future supporters who share this program's public-benefit mission.