We build ultra-long-context genomic foundation models to predict disease risk, explain heritability, and design actionable edits from whole-genome and multi-omics data.
Ultra-long context architectures pretrained on whole genomes to capture regulatory logic across megabases.
Variant effect, expression, and disease-risk prediction grounded in multi-omics, calibrated to clinical use.
From hypothesis to edit — close the loop between in-silico prediction and wet-lab validation faster.
A Nature GWAS study found a GLP1R variant linked to greater weight loss. Genomic Intelligence models suggest a second mechanism, help profile rare variants missed by GWAS, and let anyone screen their own genome.
Read →Biomni can now call Genomic Intelligence sequence-to-function models inside natural-language research workflows — turning promoter design, expression prediction, and variant interpretation into agent-executable tasks.
Read →A Claude Code agent used the Genomic Intelligence expression API to screen endogenous promoters, optimize a PROM1-derived regulatory sequence, and produce a reproducible in-silico candidate for wet-lab validation.
Read →Product vision and incremental experimentation.
Wet-lab validation and genomic model science.
Ultra-long context models and memory architectures.
Data platforms, infrastructure, and scale.
AI-for-health strategy and ecosystem leadership (Microsoft CSO; founder of Stanford AI100; former AAAI President).
AI infrastructure and large-scale ML leadership across scientific and health applications (VP & Field CTO, ML at Cerebras Systems; former Software and AI research leader at Hewlett Packard Labs).
Investors, partners, and researchers — let's talk about programming biology together.