Why we exist
Built to close the gap between preclinical signal and Phase II outcome
The attrition problem in Phase II is not primarily a compound quality problem. Most late-stage failures trace back to trial design decisions made before the IND: who gets enrolled, which biomarker threshold defines eligibility, whether the trial population is heterogeneous enough to mask a real signal in aggregate.
We built Valinor Discovery to give development teams a quantitative basis for those decisions before they are locked into the protocol. Not as a prediction service, not as a consulting engagement, but as a modeling platform that scales across programs and gives teams a reproducible method for thinking about population heterogeneity early.
The team
Three backgrounds that have to be in the same room
Translational biology, computational infrastructure, and clinical operations. Each of these perspectives shapes every decision we make about what the platform can and cannot claim to do.
Josh's background is in translational development operations, with time spent running Phase I and II programs across oncology and immunology at a pharmaceutical organization before co-founding Valinor Discovery. The hands-on experience of watching well-designed compounds fail because of trial population decisions shaped the problem the platform addresses. He oversees commercial development and program partnerships.
Priya's research background spans Bayesian graphical modeling and probabilistic simulation, with her doctoral and postdoctoral work focused on multi-modal cohort reconstruction from real-world evidence. She built the core inference engine and response surface architecture that Valinor Discovery's platform runs on, and leads the technical and data science organization.
Marcus is a translational biologist who spent his research career studying population-level biomarker variability in chronic inflammatory disease and oncology. He bridges the scientific validation requirements of pharmaceutical development teams and the computational architecture of the platform, and leads the scientific advisory and literature calibration process.
The thinking behind the platform
Why a modeling platform, not a consulting firm
The first version of this approach was a service: we would take a sponsor's cohort data and hand-deliver a report on subgroup risk. That version had a ceiling. Every engagement was custom, every output format was slightly different, and the methodology wasn't reproducible across programs by the sponsor's team independently.
The decision to build a platform rather than a consulting practice was not about scale in the conventional sense. It was about epistemic discipline: a platform forces you to commit to a methodology, document it, and stand behind it across every program that runs through it. A consulting firm can hedge. A platform cannot.
That commitment to methodological discipline is visible in how we handle uncertainty. Every output report from the platform includes confidence intervals on the key estimates. We do not report point estimates without bounds. We do not summarize outputs as definitive recommendations without qualification. The development team knows what the model's uncertainty is, and they make the decision.
We built the validation framework before we opened the platform to external programs. That meant running retrospective analyses against trials with known outcomes and measuring how well the model's subgroup failure predictions aligned with post-hoc subgroup analysis. The published results from that work are on the Science page. We are not finished validating; we expect the validation effort to continue as the platform scales.
Where we work
Boston, MA
We are based in Boston's Seaport district, close to the concentration of pharmaceutical development and biotech programs that make it one of the most active Phase I-II corridors in the world. That proximity is practical: the teams we work with are often reachable in the same day.
Boston, MA 02210
Working with us
We take on a limited number of programs at any time to maintain the quality of the analysis we can deliver. If you are evaluating the platform for a current development decision, the best first step is a direct conversation with the team.
Request a conversation