Case Study
Revolutionizing Drug Discovery with SwarmOne
Numenos leveraged SwarmOne to build groundbreaking AI foundational models using non-standard data types, dramatically accelerating the drug discovery process and launching their service significantly faster than traditional infrastructure would allow.
“SwarmOne boosted personnel efficiency by about 90%, significantly reduced training costs, and enhanced delivery, making us far more competitive in our market.”
Company
Numenos
Industry
Drug Development
Previous Infrastructure
On-premise
The Challenge
Building Novel AI for Drug Discovery at Scale
Numenos aimed to fundamentally rethink how AI could be utilized in drug discovery by testing and building an entirely new set of foundational models. This required evaluating 16,000 different model architectures using a non-standard data type - an undertaking that would traditionally demand configuring 16,000 GPU instances, a process that would have taken an impractically long time.
- Testing and building an entirely new set of foundational models for drug discovery using a non-standard data type - an undertaking without existing playbooks
- Evaluating 16,000 different model architectures, which would traditionally require configuring 16,000 GPU instances individually - an impractically long process
- No existing infrastructure could handle the scale and variety of experimental workloads simultaneously
- Need for rapid hyperparameter optimization across thousands of model variants without infrastructure bottlenecks
“I don't know how we would have done it without SwarmOne.”
The Solution
Why Numenos Chose SwarmOne
“We do a lot of experimentation. We needed a way to test thousands of model architectures and then optimize hyperparameters. We reviewed multiple options - SwarmOne was the only suite that met our needs.”
- Fully automated training: SwarmOne provisions the necessary hardware and determines the optimal training strategy - no manual GPU configuration required
- Seamless experimentation: rapid iteration across 16,000 model architectures without dealing with infrastructure overhead
- Massive scalability: handles thousands of concurrent training runs effortlessly, orchestrating GPU allocation autonomously
- Git-based code management: models managed like any other software, enabling version control and reproducibility across experiments
The Impact
Results That Speak for Themselves
10x
Trial Significance Increase
80%
Reduction in Trial Size
$200-250K
Annual Savings
16,000
Model Architectures Tested
Accelerated trials: Faster and more accurate trial outcomes with 10x increase in trial significance.
Biomarker-driven discovery:Enhanced sample prioritization and target identification through Numenos' unique Causal foundation model.
Cost savings: $200K–$250K saved annually by eliminating the need for a dedicated DevOps engineer and always-on GPU instances.
The Future
Driving Transformative Results in Drug Development
By using SwarmOne, Numenos has shifted its focus from infrastructure management to AI innovation. Their unique Causal foundation model continues to uncover data potential, driving transformative results in drug development.
In October 2024, Numenos secured investment from BC Growth Equity, a Florida-based VC firm supporting early-stage companies with exponential growth potential - a milestone enabled in part by the rapid progress SwarmOne's suite made possible.
Experience SwarmOne Today
Schedule a demo and see how SwarmOne can transform your AI infrastructure.