Building Reliable Alert Systems for AI Model Monitoring

Explore a 24-minute conference talk from Conf42 Chaos Engineering 2025 where Muddassar Sharif delves into the critical aspects of creating reliable alert systems for AI models in production. Learn about essential model monitoring techniques, key performance indicators, and how to detect data and concept drift that can impact model performance. Discover various tools for effectively monitoring AI systems, with practical examples of different drift types and their implications. Understand why tracking both input and output is crucial for maintaining model reliability, and gain insights into industry best practices and tools that can help implement robust monitoring solutions. The presentation provides a comprehensive overview from introduction to conclusion, making it valuable for engineers and data scientists working with AI systems in production environments.