What challenges did Recursion Pharmaceuticals face before AI implementation?

Drug discovery traditionally is slow and expensive, especially for rare diseases, with very few approved treatments. Recursion aimed to accelerate discovery processes, reduce costs, and improve computational speeds for data-heavy biology experiments.

What was the impact of the AI solution?

Significantly reduced model training times, enabling researchers to get results in minutes or hours instead of days. Accelerated identification of drug candidates, with some progressing to Phase 1 clinical trials. Improved platform scalability, operational efficiency, and the ability to handle large biomedical datasets. Potentially reduces time and cost in drug discovery, which traditionally takes many years.

Source

GCP

Recursion Pharmaceuticals Accelerates Drug Discovery with AI and Google Cloud TPU

Use case typeDrug discoveryUpdated Jun 13, 2026

Recursion Pharmaceuticals utilizes a hybrid lab-to-cloud platform combining biology, AI, and cloud computing to accelerate drug discovery for rare and common diseases. They leverage Google Kubernetes Engine for scalable container orchestration, Cloud TPU pods for faster deep learning model training, TensorFlow for neural network training, and Cloud Storage for data. The platform integrates experimental biology with AI pattern recognition to analyze cellular images, leading to identification of promising drug candidates.

Organization: Recursion Pharmaceuticals
Industry: Pharma
Location: United States

Reported outcomes

Strategic outcomes

Speed & agilityAccelerated drug discovery workflowsNew product / capabilityBuilt hybrid lab-to-cloud platformScale & capacityImproved large-scale biomedical data handlingNew product / capabilityAdvanced AI-driven candidate identification

Primary read

Use case focus

Showing 3 of 3

1Drug Discovery Acceleration
2High Performance Computing
3AI in Life Sciences

Drug discovery traditionally is slow and expensive, especially for rare diseases, with very few approved treatments.
Recursion aimed to accelerate discovery processes, reduce costs, and improve computational speeds for data-heavy biology experiments.

Implemented a hybrid local and cloud solution using Google Cloud technology including Kubernetes, Cloud TPU, and TensorFlow.
Migrated AI model training from local GPUs to Cloud TPU pods, reducing model training duration from over 24 hours to between 7 hours and 15 minutes.
Utilized a data processing pipeline with Kubernetes Engine and Confluent Kafka to handle large-scale microscopy image data and deep learning tasks.

Technologies

Google Kubernetes Engine Cloud TPU TensorFlow Cloud Storage Confluent Kafka

Significantly reduced model training times, enabling researchers to get results in minutes or hours instead of days.
Accelerated identification of drug candidates, with some progressing to Phase 1 clinical trials.
Improved platform scalability, operational efficiency, and the ability to handle large biomedical datasets.
Potentially reduces time and cost in drug discovery, which traditionally takes many years.

Architecture

Hybrid lab-to-cloud platform integrating local Kubernetes clusters and Google Cloud Kubernetes Engine for scalable processing; use of Cloud TPU pods for accelerated training; TensorFlow for deep learning; Confluent Kafka for data streaming.

Sources & evidence1

Groundedness: 4/5

Primary source

AI-generated summary. Verify important details with the linked sources before relying on this case.

Explore related AI use cases

Customer evidence