MLOps Framework Solution for ML Business Problems

Prior Situation / Scenario

• Business Problem: rising number of subscribers experiencing involuntary churn.
• Over 50% of machine learning projects never reach production due to labor-intensive and inconsistent workflows.
• Manual improvements of ML models are made in response to performance deterioration.
• Lack of scalability and a high impact of staff turnover because of dependency on key personnel.

Client Challenges

• Estimate the daily likelihood of involuntary churn (score/prediction) for each customer to improve Retention & Collection campaigns.
• Manage the lifecycle of the ML solution.
• Standardize operational processes.
• Enable or disable functionality remotely without deploying new code.
• Use automatic and self-healing ML models (retraining).
• Address misuse of talent: a high volume of manual tasks and low business-impact tasks.

Strata Solution/ Key Enablers

• High-performance, leading-edge ML algorithm portfolio
• ML Solution as a Product
  • Deployed in the customer cloud account
  • ML Model Development and Operationalization Manual
• Serverless and Event-Driven System
• Agile development: MVP + incremental functionalities

Outcome

MLOps Framework:

• Highly scalable and available architecture
• Delivered as a product with capabilities to: Data Processing, Prediction, Monitoring, Automatic Retraining (model optimization).
• Fast time to market with one-click deployment to production.
• Supports agile development through granular releases into production (MVP).
• Standardized, easy-to-control, operation, and maintenance​ of ML solutions.