Valinor trains virtual patient models
to make human biology predictable

Disease is found late. Drugs are tested blind. Patients are treated by population, not by biology.

What healthcare needs is a model of the patient.

Valinor is building it.

We collect more than single-cell readouts from isolated experiments — we generate virtual patient profiles grounded in rich, longitudinal data across a broad spectrum of multi-omics and clinical assays.

scRNAseq
Bulk RNAseq
Proteomics
cfDNA Methylation
Whole Genome Sequencing
Longitudinal Health Records
MRI Imaging
Amyloid PET Imaging
Histopathology

Proprietary Data Engine

Abstract circular graphic with a solid center circle surrounded by a ring of white block segments.

We generate proprietary, patient-derived datasets collected across the timepoints that matter, giving machine learning the data it needs to truly impact drug development.

Virtual Patient Representations

Three vertical blue dots circled, centered among three rows of white square icons on black background.

We learn representations of individual patients from multimodal longitudinal data across disease and treatment.

Interpretability

Three red and three pink fuzzy spheres arranged diagonally on a black background.

We embed DNA, methylation, transcriptomics, proteomics, imaging, and structured clinical data into one space, capturing the heterogeneity of patient biology.

We train virtual patient models in diseases where patient biology is least understood.

Valinor is backed by CRV, Mythos Ventures, Pelion Venture Partners, Harpoon Ventures, Amino Collective, and founders from Anthropic, Goodfire, and Mercor.