United StatesHealthcare and life sciences2023 – ongoingVertex Pharmaceuticals
Parallel ML image segmentation for scientific experiments
AWS reported up to 11 times faster image segmentation at Vertex after parallelizing processing
Problem
Scientists' microscope images queued on shared computing resources. Heavy demand delayed results, while idle capacity added cost.
Approach
Scientists labeled sample images to fine-tune a proprietary segmentation model, then applied it to the remaining images. Vertex redesigned the workflow with AWS. Step Functions Distributed Map and Lambda ran image processing in parallel.
Results
Takeaway
The reported speed and cost gains came from reorganizing an existing ML workload. They did not measure improved model accuracy or drug-discovery success.
Categories
- Industry
- Healthcare and life sciences
- Business function
- IT and R&D
- Innovation type
- Process innovation
- AI technology
- Computer vision
- Adoption
- Partner co-development
- Evidence
- Vendor case study
Sources
Source summaries
AWS and Vertex engineers described the deployed ML segmentation workflow. They reported jobs completing up to 11 times faster and a 90% cost reduction after parallelizing processing.
Full details are in the original sources.