AI case study

RecursionDrug discovery

Physics simulations took weeks to screen libraries. Boltz-2 runs 1,000x faster, cutting this to hours with 2x precision.

Published|4 months ago

Key results

Processing Speed
1,000x
vs physics pipelines
Wet Lab Work Required
40%
Precision Improvement
2x
vs docking and prior ML approaches

Result highlights

Unlock 3 result highlights

The story

Context

A clinical-stage biotechnology company manages 50 petabytes of biological and chemical data to industrialize the drug discovery process.

Challenge

Traditional physics-based simulations required weeks to screen compound libraries, while earlier AI models could not accurately predict how strongly...

Solution
Unlock full story

The company

Recursion logo

Recursion

recursion.com

AI-driven drug discovery and clinical-stage biotechnology platform.

IndustryPharmaceuticals & Biotech
LocationSalt Lake City, UT, USA
Employees251-1K
Founded2013

The implementation partner

Massachusetts Institute of Technology logo

Massachusetts Institute of Technology

web.mit.edu
Role in this case study

MIT researchers co-developed the Boltz-2 biomolecular foundation model with Recursion.

IndustryEducation & Training
LocationCambridge, MA, USA
Employees10K-50K
Founded1861

The vendor

NVIDIA is a technology company that specializes in semiconductors, graphics processing units, and artificial intelligence for applications in data centers, gaming, and more.

IndustryTechnology
LocationSanta Clara, California, United States
Employees10K-50K
Founded1993

Use case

Recursion's Drug discovery is part of this use case:

Scientific Discovery
21 case studies(+150% YoY)
Proven impact?
LowModerateVery Strong
4.1Moderate
4.3Moderatewithin Pharmaceuticals & Biotech
3.9Moderatewithin Product Engineering

Similar Case Studies

Related implementations across industries and use cases

Phagos logoP

Phagos

Pharmaceuticals & Biotech|SMB

Drug discovery

Matching phages required screening trillions of options. Gen AI simulates interactions, cutting discovery from 10 years to 2 months.

2 monthsDevelopment Timevs 10+ years traditional development
via aws.amazon.com
Published Dec 6, 2025
Harvard Medical School logoH

Harvard Medical School

Pharmaceuticals & Biotech|Enterprise

Protein interaction discovery

Mapping interactions took years. A custom AI pipeline analyzed 1.7M predictions in 3 months, finding 40,000 high-confidence pairs.

3 monthsProject Timelinevs years of manual work
via nvidia.com
Published Jan 30, 2026
Retro Biosciences logoR

Retro Biosciences

Pharmaceuticals & Biotech|SMB

Protein engineering

Navigating 10^1000 variants was impossible. A model designed novel proteins, boosting stem cell marker expression >50x over controls.

>50xMarker Expressionvs wild-type controls
via openai.com
Published Aug 22, 2025

23 AI case studies in Scientific Discovery

91 AI case studies in Pharmaceuticals & Biotech

574 AI case studies in Product Engineering