We make the microbiome computable
Anto is a frontier AI research lab based in San Francisco.
We are researchers building multimodal foundation models for microbial communities. For the first time, we can computationally model how the microbiome affects drug response in individual patients. This enables precision medicine that accounts for the bacterial ecosystems inside us, not just our human genome.
Arvid E. Gollwitzer
Nature-published researcher who pioneered microbiome sparsification. He built the methods that make noisy microbial data tractable, enabling foundation-model pretraining on real-world corpora.
Arvid pioneered computational sparsification: training systems to ignore 99.9% of noise and focus on the 0.1% signal that drives outcomes. He is the creator of GenStore, MegIS, and MetaTrinity, achieving 10-100× performance gains in genomics through hardware/software co-design.
A second-time founder (previously NextGuide, AI for the visually impaired), his work has been featured in ISCA, ASPLOS, NeurIPS, and Nature. He is a Swiss Study Foundation fellow (top 0.3%) and teaches Clinical Genomics at ETH Zürich.
David de Gruijl
David was a researcher at Brigham and Women's Hospital and Harvard Medical School, where he built in-vitro gut models to study microbial interactions.
Previously, he worked in data science at Johnson & Johnson and founded an AI startup for pharmacokinetics at ETH Zürich. David saw firsthand how broken drug-microbiome testing is in current clinical workflows.
He specializes in translating bench science to computational models, knowing exactly where current workflows break and how to fix them to address the microbiome-driven causes of drug response and failures.
Making sense of the microbial world
Traditional machine learning fails on microbiome data because it treats all features equally. Our goal-directed sparsification learns which microbial features actually matter for specific outcomes. By focusing computational resources on signal rather than noise, we achieve both better predictions and mechanistic interpretability.
The result is a foundation model that understands microbial ecosystems at a fundamental level—one that can predict drug-microbiome interactions, identify biomarkers, and guide therapeutic development.
.avif)
