Skip to main content

Agenda

Loading

Environment Agnostic ML: Achieving Portability Without Sacrificing Performance

16 May 2024
People X Leadership Stage
AI , Technical

Development of machine learning tends to get code integrated with the underlying infrastructure and platforms resulting in restrictions to portability. Yet with proper methods, you can disintegrate your ML pipeline, so that it can run anywhere without losing fullness.

Techniques and best practices such as abstraction layers, containerization, cloud-agnostic APIs, and automated testing that facilitate environment-agnostic ML will be discussed in this talk. You will find out an architectural approach and tooling that enables fast porting models to be moved from on-prem, cloud to a hybrid environment.

We will also delve into the optimization strategies of performance, such as dynamic allocation, profiling, and tuning and not be faced with trade-offs when moving among environments. You will end up knowing how to construct supportive, tactile ML systems that give delivery mobility from model to way of life. Disengage your ML pipeline and code from platform limitations without compromising efficiency or timeliness.

Key Takeaways:

  • Decouple your ML code from specific environments through modular, containerized architectures and platform abstraction layers. This prevents locking into any one platform.
  • Leverage cloud-agnostic APIs and libraries with fallback support for different platforms rather than platform-specific tools. Minimizes rework when porting.
  • Automate performance testing across target deployment platforms early and often. Enables tuning models for optimal performance everywhere.
Speakers
Mireille Estephan, Chief Technology Officer - Ardent
View all Agenda
Pre-footer CTA

Pre-footer CTA