New Way Now: Recursion processes petabytes of data into therapeutic breakthroughs with Google Cloud

*Featured in this video:* Maureen Makes, VP of Engineering at Recursion *Executive summary:* Recursion, a leading biotech company, partnered with Google Cloud to develop the Recursion OS, a drug discovery and development AI platform that connects lab experimentation with in silico analysis. The platform processes more than 65 petabytes of data and runs up to 2.2 million experiments a week, using powerful AI models to analyze hundreds of millions of cellular images and map biological and chemical relationships that can be tested in the lab. The Recursion OS has already improved the cost and timeline of discovery, pushing some target discovery programs through to preclinical studies in just 18 months and cutting the cost of model inference by as much as 50%. Recursion is also leveraging generative AI, including Google’s Gemini models, to accelerate and automate tasks across their organization. This allows their team to focus on solving bigger, more strategic problems while the AI helps with things like improving search and access, speeding up high-quality document creation, and resolving common IT issues with the help of an AI agent. *Challenge:* Traditional approaches to drug discovery are notoriously slow, expensive, and inefficient, limiting the number of new therapies available to patients. Driven by its mission to decode biology, Recursion wanted to industrialize the drug discovery process, boosting the speed and success rate of identifying viable therapeutic compounds while simultaneously lowering the overall cost of discovery. *Solution:* Google Cloud forms the core foundation of Recursion OS; Cloud Storage provides cost-effective, secure storage for over 65 petabytes of proprietary data, BigQuery makes it easy to manage and analyze these massive datasets, and Looker simplifies data access for non-technical users. With AI-optimized infrastructure like Cloud TPUs, Recursion has the ability to access and scale up resources to run model inference at scale for