Building an enterprise MLOps platform for a major bank

At a glance

One of Australia’s big 4 banks engaged Mantel to build an enterprise-grade, multi-tenant MLOps platform to standardise AI/ML lifecycles, upskill internal teams, and support the scaling of high-value models. This initiative was key to supporting our client’s 100+ models for critical business decisions, particularly in risk management and lending.

Key outcomes
  • Accelerated Model Development and delivery
  • Foundation for a scalable AI Factory
  • Standardised AI/ML lifecycle management
  • Upskilled internal teams to sustain and govern the new platform

The challenge

Our client’s Risk division relies on over 100 models for crucial business decision-making in lending and risk management. The bank required an enterprise-grade, multi-tenant solution to enable diverse teams to onboard and deliver high-performing models quickly while maintaining robust governance. Our client chose a leading cloud provider for its next-generation applications, and Mantel was engaged as a key transformation partner to design and build a centralised MLOps platform that resolves the balance between agility and control.

The solution

Mantel designed and built an end-to-end enterprise MLOps platform on our client’s chosen Cloud Platform, including patterns for productionisation and deployment. The solution delivered included:

  • Designed and built MLOps architecture and multi-tenant environments across the entire Risk division.
  • Implemented a key cloud service for standardised, regulated lifecycle management of AI/ML models.
  • Established a reusable framework with 14 standardised, auditable pipelines for high-value retail models, adhering to enterprise and regulatory standards.
  • Executed a co-delivery model and upskilled staff to ensure internal teams could sustain and govern the new enterprise capability.

The outcome

The capability uplift, built platform, and standardised process successfully delivered for our client’s Risk division the ability to:

  • Accelerate Model Development: The bank’s most complex, high-value retail models were delivered with rapid feedback loops and multiple iterations under stronger governance and security.
  • Enable Enterprise Scale: Multi-tenancy and repeatable blueprints formed the foundation of an AI Factory for faster, scalable delivery.
  • Adopt, Operate & Scale the Platform: Teams adopted and maintained the platform while cascading knowledge and embedding new ways of working.