Living Models, a Paris–Berkeley
startup, has raised $7 million in seed funding as it emerges from stealth to
develop foundation models for biology trained on DNA, RNA, and multi-omics data
aimed at improving understanding of biological systems.
To support the next
stage of development, the company has also secured access to a computing
cluster of 120 NVIDIA B200 GPUs, which it plans to use to train its next
generation of biological AI models.
The company develops large-scale
transformer models trained on genomic, transcriptomic, and other biological
datasets to analyse patterns within living organisms. Operating in Paris and Berkeley, Living Models brings together researchers in artificial
intelligence and plant science to apply machine learning to biological research
and agricultural innovation.
While artificial intelligence has
already transformed sectors such as finance, software development, and content
creation, its application in areas such as agriculture and food production
remains at an earlier stage.
Living Models is focusing on this area
by applying AI techniques to biological data, particularly in plant science,
where improving crop resilience and productivity is becoming increasingly
important as climate pressures affect global agriculture.
As part of its launch, the company
introduced BOTANIC, a family of transformer models designed for plant biology.
The models are trained on genomic sequences from multiple plant species and analyse genomic and other biological data to identify genetic markers associated
with traits such as climate resilience and disease resistance.
By predicting
which genetic variants are worth testing, the technology aims to help seed and
agricultural companies accelerate the development of new crop varieties.
OpenAI trains on Reddit and Wikipedia
to understand human language. We train on DNA, RNA, and gene expression to
understand the language of life itself,
said Cyril Véran, CEO and co-founder
of Living Models.
Traditional crop breeding cycles can
take many years, partly due to the time required to identify promising genetic
traits. By analysing genomic data computationally, Living Models aims to
shorten the early stages of this process by helping researchers focus on the
most relevant genetic variants before conducting field validation.
In
the longer term, Living Models plans to expand its work on foundation models
for biological systems beyond plants. The company began with plant biology due
to the availability of large genomic datasets, faster validation cycles
compared with other life-science fields, and the growing need for technologies
that support climate-resilient agriculture.