FMCG

Standardizing FMCG Operations with the MLOps best practices Across Multiple Cloud Platforms

Client

Global FMCG / CPG Company

Date

Services

MLOps

Technologies

Python, Docker, Kubernetes, Azure, GCP, Databricks, ChatGPT, Langchain, CI/CD with GitHub Action

Challenge

The client faced difficulties in scaling and expanding various ML use cases across multiple cloud platforms. They required a solution to standardize their operations, streamline ML training and inference tasks, and ensure flexibility and scalability across different environments.

Our approach

Our team undertook a transformative project to standardize the client’s ML operations using MLOps templates and best practices. By leveraging cloud technologies and AI, we developed a cloud-agnostic tool and a customized MLOps template that seamlessly integrated ML use cases across GCP, Azure, and Databricks.

Key components of the solution included:

  • Automated CI/CD processes aligned with organizational standards.
  • Resource autoscaling and consumption logging to optimize cloud resources.
  • Model versioning, monitoring, and retraining pipelines, triggered by data or concept drift detection.
  • Deployment of a Q&A chatbot powered by ChatGPT and Langchain, offering real-time support for data scientists navigating the tool and templates.

The outcome

The cloud-agnostic tool and standardized MLOps template brought significant improvements to the client’s ML operations. Maintenance tasks were streamlined, reducing overhead costs and increasing operational efficiency. CI/CD pipelines implemented with GitHub Actions accelerated the development cycle, enabling faster transitions from development to production.

Additionally, the integration of a chatbot provided users with real-time support and guidance, boosting user satisfaction and productivity.