AI and deep learning integration into research methodologies is not just a trend but a necessity. Every area of scientific research now has a large component of high-resolution data that is impossible to make sense of without appropriate computational tools.

Talmo Pereira
Salk Fellow and Principal Investigator, Salk Institute
Overview

Nestled along the Pacific coastline, the iconic buildings of the Salk Institute for Biological Studies have housed scientific visionaries and trailblazers for over 60 years. The Institute is dedicated to improving the quality of life worldwide with advancements in neuroscience, cancer research, aging, immunobiology, plant biology, computational biology, and more.

Behind these life-changing innovations lies an organization deeply committed to pushing the boundaries of human knowledge through pioneering research fueled by ingenuity and enabled by cutting-edge infrastructure. One innovator at the frontier is Talmo Pereira, Salk Fellow and Principal Investigator. Today, Talmo focuses on quantifying motion using deep learning and computer vision to discover behavioral biomarkers of disease - understanding how movement patterns indicate disease.

Background

Even trailblazing institutes must periodically modernize systems and workflows to stay at the top. Robust data pipelines and shared AI/ML infrastructure empower teams to unlock AI’s full potential.

Recognizing AI/ML’s broad potential and critical need across research disciplines, Talmo embarked on an ambitious BioComputing Initiative to evolve the scientific and research computing infrastructure. Continuing pioneering work required an infrastructure supporting data-intensive artificial intelligence workloads beyond what existing systems delivered.

“AI and deep learning integration into research methodologies is not just a trend but a necessity. Every area of scientific research now has a large component of high-resolution data that is impossible to make sense of without appropriate computational tools,” explains Talmo. Through his lab’s specialization in applying deep learning and computer vision to study biology computationally, Talmo witnessed firsthand AI/ML’s transformative impact.

Outcome

In reimagining its research data architecture, the Salk Institute turned to VAST Data, the industry-leading platform for AI-powered computing. “We wanted to future-proof the Institute’s data capabilities,” Talmo notes. After conducting in-depth due diligence evaluating solutions on performance, scalability, and total cost of ownership, VAST Data emerged as the clear choice. Talmo’s lab runs on premier infrastructure comprising many AI/ML enabling technologies, including VAST, NVIDIA, and Run.ai. Run.ai offers fair-share scheduling, allowing the Salk data science and AI team to quickly and automatically share clusters of GPUs without memory overflows or processing clashes.

The VAST Data platform rollout has delivered rapid results. Talmo observes tangible improvements in his lab’s AI/ML pipelines, highlighting noticeable latency reductions in streaming research image and video datasets from centralized data tiers to GPU compute nodes. This accelerated data ingestion enables faster model development cycles.

The increased throughput is a huge help for our deep learning workloads where high I/O is critical,” he explains. “Being able to mount and operate seamlessly across platforms without file permission or flag issues makes our lives much easier. There’s zero IT overhead regarding management, since VAST handles that for us.

By taking a strategic, forward-looking approach to modernizing its research data infrastructure, the pioneering Salk Institute continues pushing boundaries - poised to accelerate the next wave of scientific breakthroughs.