Valent U.S.A. LLC

Serverless Machine Learning

• Using Serverless for MLops allows Valent data scientists to push new models to dev, staging, or production.
• This means that data scientists are the owners of product reliability, ensuring that they maintain and guarantee the quality of their model’s results.
• Turn around times are dramatically reduced using this architecture, as data scientists own the full data pipeline from storage, through feature engineering, model building/validation/tuning, and out the door to customers. There is no more waiting for outside teams to deliver value to customers.

Managed Services Instead of Manual Services

• Using cloud-provider managed services for API access through automated gateway deployment, computation through serverless cloud functions, and CI/CD through hosted integration and deployment pipelines streamlines ML deployment so that data scientists spend their time optimizing models, not doing data engineering.
• Deploying the entire web app consuming the results of the ML models using the same framework means that all services operate in the same isolated cloud environment. This fully isolates the dev, staging, and production environments for consistency and reliability.