A stealth Series B biotech startup focused on developing innovative therapies faced significant challenges in managing its cloud-based high-performance computing (HPC) workflows. The company’s computational workloads included high-throughput virtual screening (HTVS), molecular docking, molecular dynamics simulations, and AI-augmented HTVS; each requiring immense computational power. However, managing cloud infrastructure, optimizing HPC strategies, and scaling clusters created bottlenecks that slowed the team’s research and increased operational costs.
Fovus stepped in to revolutionize the startup’s approach, streamlining cloud logistics and optimizing HPC strategies with AI-powered automation. This allowed the startup to accelerate drug discovery timelines, cut cloud costs, and free up valuable resources for scientific innovation.
“Fovus supercharged our computational drug discovery pipelines. Fovus accelerated our time-to-insight in DMTA cycles by 96x, at one-fifth of the overall cloud cost.” – Director of Computational Chemistry
The Challenge: Navigating Cloud Complexity and Cost Inefficiencies
The startup’s team spent too much time and effort managing cloud infrastructure, diverting focus from research and development. Several key challenges were slowing progress:
Managing cloud logistics consumed 10-20% of the team’s time, creating innovation downtime. Cloud offerings were diverse and highly heterogeneous, making it challenging to identify the most efficient HPC strategies for different workloads. Strategies that worked well for one workload often performed poorly for another, resulting in inefficiencies and higher costs. Scaling up cluster sizes for large workloads, particularly GPU-intensive tasks, was difficult due to a lack of cloud expertise, leading to longer time-to-insight and slower DMTA (Design, Make, Test, Analyze) cycles.
Without a solution to optimize these processes, the startup faced rising cloud expenses and slower drug discovery timelines—two challenges that needed immediate attention.
How Fovus Solved These Challenges
Fovus provided a comprehensive solution by automating cloud management and optimizing HPC strategies with AI, allowing the biotech team to focus on advancing their research.
Severless and Effortless HPC Orchestration with AI
Fovus seamlessly orchestrated cloud logistics and optimized HPC strategies using AI, allowing the team to say goodbye to cloud management headaches. With an intuitive web UI, CLI, and Python API, tapping into supercomputing power became as easy as running a command or clicking a button.
Free Automated Benchmarking for Optimized Workloads
Fovus auto-benchmarked the startup’s computational chemistry workloads to identify optimal CPU, GPU, memory, and storage configurations. This custom benchmarking data became the foundation for fine-tuning HPC strategies for each workload, ensuring maximum performance at the lowest possible cost.
AI-Optimized HPC Strategies to Minimize Cost and Time
By leveraging custom benchmarking data, Fovus automatically determined the most efficient HPC strategy to minimize runtime and cloud costs for every workload. This eliminated the need for trial-and-error tuning, saving both time and money.
Multi-Cloud Region Auto-Scaling for Massive Parallelization
Fovus auto-scaled clusters across multiple cloud regions and availability zones to maximize resource availability and accelerate computational tasks. This approach enabled the team to run large-scale HTVS workloads in parallel, reducing time-to-insight by nearly 100x.
Continuous HPC Strategy Improvement
Fovus didn’t just optimize once—it continuously updated HPC strategies as new cloud infrastructure and hardware technologies became available. For example:
Initial optimization delivered 1.6x faster performance and 2x lower costs using ARM CPUs.
Six months later, Fovus auto-upgraded the strategy with next-gen x86 CPUs, delivering 1.5x faster performance and 1.3x lower costs.
Intelligent Spot Instance Usage for Extra Savings
Fovus analyzed real-time spot pricing, interruption frequency, and workload runtime predictions to dynamically leverage spot instances, reducing cloud costs by an additional 2-3x. Even if a task was interrupted, Fovus auto-requeued and re-executed it, ensuring data integrity.
Transformative Results: Faster, Cheaper, and More Scalable Discovery
By leveraging Fovus, the biotech startup experienced transformational improvements in its computational drug discovery workflows:
| Workload | Self-Managed Time | Self-Managed Cost | Fovus Time | Fovus Cost | Time Reduction | Cost Savings |
|---|---|---|---|---|---|---|
| HTVS (Docking) | 45 hrs | $3,250 | 3.5 hrs | $650 | 13x | 5x |
| AI-augmented HTVS (Prediction) | 48 hrs | $800 | 0.5 hrs | $100 | 96x | 8x |
| Total (5 DMTA Cycles) | 24 days | $20,250 | 6 days | $3,750 | 4x | 5.5x |
For docking-based HTVS, time-to-insight and cloud costs were reduced significantly for each step. Key outcomes include:
- 13x Faster Docking-Based HTVS: Reduced time-to-insight from 2 days to 3.5 hours.
- 96x Faster AI-Augmented HTVS Predictions: Reduced time-to-insight from 2 days to 30 minutes.
- 5.5x Lower Cloud Spending: Cut total cloud costs by more than half over 5 DMTA cycles.
- 8000-Node Cluster Deployed in Under 20 Minutes: Massively parallelized workload execution across cloud regions.
Increased Productivity and Innovation
By automating cloud integration and management, Fovus allowed the biotech team to focus on their core mission: advancing drug discovery. The team reported a 15% increase in work productivity as they spent less time troubleshooting cloud infrastructure and more time on high-impact research. With faster and more cost-effective computational workflows, the startup could iterate through DMTA cycles faster, ultimately accelerating the journey from discovery to clinical validation.
AWS Partnership
Fovus is a trusted Software Partner of AWS. Fovus leveraged AWS infrastructure and services, combined with its deep experience and expertise in High-Performance Computing (HPC), to create an AI-powered serverless HPC platform.
Fovus utilizes Amazon EC2 instances optimized for HPC workloads, enhanced with Elastic Fabric Adapter (EFA) to enable low-latency, high-speed interconnect between compute nodes. Amazon EBS and Amazon EFS provide persistent and shared storage, ensuring fast access to data. Data is stored securely using Amazon S3, with Amazon S3 Glacier enabling cost-effective long-term archival. To further accelerate data-intensive workflows, Amazon FSx for Lustre is integrated, delivering high-throughput file access.
Unlock the Future of Computational Drug Discovery with Fovus
Fovus empowers biotech companies to scale their computational workloads efficiently while reducing costs and accelerating time-to-insight. Discover how Fovus can optimize your computational workflows and streamline your HPC strategies.