Get 20% off all career paths from fullstack to AI
AI Engineer - Learn how to integrate AI into software applications
Overview
AI, Data Science & Cloud Certificates from Google, IBM & Meta — 40% Off
One plan covers every Professional Certificate on Coursera. 40% off Coursera Plus Annual.
Unlock All Certificates
Explore the next generation of observability in this 38-minute technical talk where Neel demonstrates how modern monitoring is evolving beyond traditional logs, metrics, and traces into a predictive, AI-powered discipline. Learn about the transformation of observability through OpenTelemetry, machine learning, and large language models, starting with the evolution from basic monitoring to sophisticated predictive systems. Discover dynamic sampling techniques that significantly reduce operational costs while maintaining comprehensive system visibility, and understand how machine learning algorithms can detect anomalies and predict system failures before they impact customers. Examine practical implementations using tools like the OpenTelemetry Collector, with real-world scenarios covering everything from managing massive log volumes to implementing cost optimization strategies. Delve into LLM observability concepts, context windows, and token management, while exploring the potential for self-healing systems and understanding the edge cases where dynamic sampling approaches may not be suitable. Gain insights into cost optimization strategies and learn when and how to implement these advanced observability techniques in production environments.
Syllabus
Welcome & Introduction
Neel's Background & Community Work
The Evolution of Observability
6:29 The 2 AM Production Incident Scenario
OpenTelemetry's Role in Modern Observability
Dynamic Sampling Techniques
ML & AI in Anomaly Detection
LLM Observability Explained
Cost Optimization Strategies
Context Windows & Token Management
Self-Healing Systems Discussion
Edge Cases: When Dynamic Sampling Doesn't Work
Wrap-up & Resources
Taught by
vBrownBag