Understand how Reveal’s MLOps platform helped a leading pharmaceutical company train, deploy, run, manage, and monitor hundreds of machine learning (ML) models to facilitate drug discovery.
Implementing Multiple ML Models at Scale with Existing Infrastructure
Our client wanted to leverage ML models to aid them in drug discovery. However, ML models need a large number of labeled datasets to learn from. Additionally, deploying these ML models on the available infrastructure was a huge load for the client.
Reveal’s MLOps Solution
After understanding the client’s challenges in detail, Reveal developed a custom MLOps platform for them. The platform leveraged off-the-shelf components such as SageMaker, MLFlow, and Grafana with appropriate tweaks to support the unique use case. It was able to train, deploy, run, manage, and monitor hundreds of ML models at scale. It was built with APIs that had appropriate access controls to support the needs of different teams in the organization.
Positive Results for the Client
Once the MLOps platform was built and implemented, the client observed the following improvements –
- Access control became fairly simple to implement helping improve efficiency across teams
- Training, deploying, running, managing and monitoring hundreds of ML models was made feasible within the client’s infrastructure at a nominal investment
- The client was able to focus on their core offering – developing new drugs – with much faster turnaround cycles