Spot by NetApp: Cloud automation and optimization enhanced to deliver cost fluctuation predictions

Qwak was engaged to streamline the transition of the team's ML models to production and bridge the gap between data science and machine learning engineering teams.

About Spot by NetApp

Spot by NetApp transforms cloud operations through innovative machine learning and analytics, automating and optimizing infrastructure for optimal, scalable, and cost-effective performance. The platform maximizes cloud investments with cutting-edge technology and predictive insights into infrastructure and cost fluctuations.


SaaS, Cloud Operations

Use Case

XGBoost based on Elasticsearch
Linear regression based on S3 data with daily retrain

Model Frameworks

No items found.

Before Qwak, delivering a new AI model took weeks... Now the research team can work independently and deliver while keeping the engineering and product teams happy.

Spot by NetApp

Faster time to ML model deployment


Less issues in production


New models running in production within 4 weeks


  • Deliver models to production efficiently and in a timely manner
  • Enhance Data Science version deployment with less engineering and/or DevOps dependencies
  • Test model versions and feature creation without impacting production environments
  • Centrally monitor model performance metrics and control changes. Align feature data pipeline



Qwak Build: standardizes an ML delivery project structure This allows every team member to receive the same code structure, data versioning and testing capabilities for every model version.Using Qwak Build increases team members' confidence that new versions won't break any production service and ensures the ability to rollback to previous versions at any time.

Qwak Hosting: allows Spot’s team to deploy a new real-time model with a click of a button and automatically support Spot’s deployment strategy.

Model Serving: allows Spot’s team to deploy a new real-time model with a click of a button and automatically support Spot’s deployment strategy which is to deploy a canary version that only a subset of users gets as the first part.

Model Monitoring: stores all the inference (prediction) data automatically and provides the ability to query real production data at any stage.

Feature Store: gave Spot’s Data Science teams the ability to manage their own data pipeline while writing and scheduling their own features. This eliminates the dependency on engineering teams.By using the features that we built within the feature store we’re now able to train our models and serve them with ease. Moreover, Spot now have visibility of their feature freshness, lineage and one simple catalog that allows us to easily connect features to models.