Why CI/CD Fails for AI and How CC/CD Fixes It
MLOps World: Machine Learning in Production via YouTube
-
64
-
- Write review
Overview
Coursera Spring Sale
40% Off Coursera Plus Annual!
Grab it
Explore a conference talk that challenges traditional software deployment approaches for AI systems and introduces a revolutionary framework for managing artificial intelligence in production. Learn why conventional Continuous Integration/Continuous Deployment (CI/CD) pipelines fail when applied to AI systems that exhibit non-deterministic behavior and autonomous decision-making capabilities. Discover the Continuous Calibration/Continuous Development (CC/CD) framework, a new methodology specifically designed for building and scaling AI products that operate unpredictably and independently. Gain insights from industry experts who have analyzed over 50 real-world AI deployments, understanding how teams can begin with low-agency, high-control configurations and gradually scale AI systems as they demonstrate reliability and earn trust. Master the art of scoping AI capabilities and designing meaningful evaluation metrics that account for the inherent uncertainty in AI behavior. Understand how to implement effective behavioral monitoring systems that prevent model drift and maintain system performance in production environments. Acquire practical strategies for scaling AI autonomy while maintaining necessary control mechanisms, ensuring safe and reliable AI system deployment. Learn to navigate the unique challenges of AI system development, from initial evaluation design through production monitoring and trust calibration processes.
Syllabus
Why CI/CD Fails for AI & How CC/CD Fixes It | Aishwarya Reganti (LevelUp Labs) & Sai Kiriti (OpenAI)
Taught by
MLOps World: Machine Learning in Production