Bioptimus

Bioptimus is an AI biotech company developing universal foundation models for biology to integrate multi-scale, multimodal biological data. The company was founded in Paris, France, in 2024 by a team of scientists from Google DeepMind and Owkin, led by CEO Professor Jean-Philippe Vert. The founding team also includes Rodolphe Jenatton (CTO), Zelda Mariet, and Felipe Llinares from Google DeepMind, along with David Cahané and Eric Durand from Owkin. Jean-Philippe Vert is a former research lead at Google Brain and was the Chief R&D Officer at Owkin, bringing extensive experience in machine learning and life sciences.

Bioptimus launched with an initial seed funding round of $35 million in February 2024, co-led by Sofinnova Partners and Bpifrance Large Venture, with participation from other funds and investors including Owkin and Xavier Niel. Within a year, the company secured a $41 million Series A round, bringing its total funding to $76 million. This funding is allocated towards computational power, data access, and team expansion.

The company's core mission is to build AI foundation models that learn from vast and diverse biological data—from molecules and cells to tissues and whole organisms—to uncover complex biological laws. This approach aims to break down data silos that have traditionally fragmented biological research. The company's business model involves providing its models through a Software-as-a-Service (SaaS) licensing structure. This includes paid research partnerships with pharmaceutical companies to develop custom models and a pay-per-use API access for developers. Bioptimus serves clients in the pharmaceutical and biotechnology industries, with 12 of the top 20 pharmaceutical companies using its models.

Bioptimus has released several flagship models. H-Optimus-1 is a leading foundation model for histology (the study of tissues), trained on a massive dataset of pathology slides. An earlier version, H-Optimus-0, was released as an open-source model with 1.1 billion parameters. The company also unveiled M-Optimus, its first multimodal "World Model" for biology, designed to simulate biological complexity by integrating various data types like histology, multi-omics, and clinical outcomes. These models enable applications such as accelerating drug discovery, designing more efficient clinical trials by creating "digital twins" of patients, predicting treatment responses, and identifying biomarkers.

Keywords: foundation models, generative AI, computational biology, multimodal data, digital pathology, histology AI, precision medicine, drug discovery, digital twins, clinical trial optimization, multi-omics data, Jean-Philippe Vert, Sofinnova Partners, Bpifrance, Owkin, Google DeepMind, H-Optimus, M-Optimus, AI for biology, biomedical research