Phagos uses generative AI on AWS to match bacteriophages to bacterial infections
Phagos · France
Phagos, a Paris-based biotech startup, uses generative AI models built with Amazon SageMaker AI to match bacteriophages to target bacteria for phage therapy, replacing a manual trial-and-error process. The AI models cut wet-lab testing needs by 50% and reduced phage-candidate screening time by 99.5%, from 29 hours to 10 minutes per bacteria. Phagos has treated more than half a million animals in France and can now develop a new treatment in two months versus 10+ years for traditional antibiotic development.
Overview
Phagos, a Paris-based biotech startup, uses generative AI models built with Amazon SageMaker AI to match bacteriophages to target bacteria for phage therapy, replacing a manual trial-and-error process. The AI models cut wet-lab testing needs by 50% and reduced phage-candidate screening time by 99.5%, from 29 hours to 10 minutes per bacteria. Phagos has treated more than half a million animals in France and can now develop a new treatment in two months versus 10+ years for traditional antibiotic development.
The challenge
Antibiotic resistance is a growing global threat, with bacterial infection expected to become the leading cause of mortality by 2050, and traditional antibiotic development can take more than 10 years and cost billions of dollars. Matching the right bacteriophage to a target bacterium had been a largely manual trial-and-error process, an exponential problem unable to be solved manually given the astronomically large number of possible phage-bacteria combinations.
The solution
Phagos uses Amazon SageMaker AI to train and fine-tune generative AI models on genomic data, from public databases and its own lab-generated data, that simulate millions of phage-bacteria interactions and determine the optimal phage characteristics for killing a given bacteria strain, powering its proprietary AI platform, Alphagos. Phagos also relies on high-performance GPU instances on Amazon EC2 for the computational power needed to run biological simulations, Amazon S3 as its central data lake, and Amazon RDS to manage and query structured data, working closely with AWS and AWS consulting partners.
Reported business value
The AI predictions reduce wet-lab testing needs by 50% and deliver a 99.5% time savings when screening phage candidates, down from 29 hours to 10 minutes per bacteria. Phagos can now develop a new treatment in two months from scratch versus 10+ years for traditional antibiotic development. Phagos has treated more than half a million animals in France after demonstrating efficacy over more than 10 clinical trials, and received approval from the French Agency for Veterinary Medicinal Products to market its solution in France as personalized veterinary medicines. In an early field test at a French oyster farm, applying an identified phage reduced oyster mortality by 40%.
Sources
Open any source and check the claim yourself — that is the point of the register.
Other life sciences entries in the register.
Novo Nordisk accelerates clinical documentation and drug development with Claude
Novo Nordisk, a global pharmaceutical company headquartered in Europe, built NovoScribe, an AI-powered documentation platform using Claude Code, Amazon Bedrock and MongoDB Atlas, to automate clinical study report (CSR) generation. Claude helped cut writing times on CSRs by 90%, reducing time spent producing clinical study documentation from 10+ weeks to 10 minutes, and delivered a 95% reduction in resources needed to create device verification protocols. The platform expanded from clinical trial reports to device protocol documentation and patient materials, and an 11-person development team, including non-technical staff, now prototypes features using Claude Code.
Novartis: Accelerating Drug Development with AI-Powered Clinical Trial Transformation
Novartis partnered with AWS Professional Services and Accenture to modernize their drug development infrastructure and integrate AI across clinical trials, with the goal of reducing trial development cycles by at least six months. The initiative built a GXP-compliant data mesh platform on AWS with Databricks for processing, enabling AI use cases including protocol generation and an intelligent decision system (digital twin). Early results from the patient safety domain showed 72% query speed improvements, 60% storage cost reduction, and 160+ hours of manual work eliminated. The protocol generation use case achieved 83-87% acceleration in producing compliant protocols.
Was this helpful?
Your feedback helps us improve our use case database