AI Monitoring Mastery: Optimize Performance, Ensure Compliance, and Cut Costs
summary
Shows concrete monitors and measurements: request traces that break latency by span (model call vs DB vs app), token counts per request and a sample that uses 372 tokens used, cost math translating tokens to dollars, and demos wiring OpenAI, AWS Bedrock, Pinecone embeddings and a multi‑agent system that runs end‑to‑end (QR links to GitHub). Compares models by latency, error rate and tokens, surfaces compliance options (drop or obfuscate content) and highlights OpenTelemetry projects for integration.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:08AI monitoring critical today
- 5:48Key challenges in AI monitoring
- 9:22Complexity with AI apps
- 17:17Telemetry for AI quality and cost
- 22:50Security and compliance in AI
- 26:40AI monitoring solutions
- 34:23Summary and future
- 35:49Q&A
related talks
40mOptimizing AI Applications for Production with Observability
31mDeep Dive To AI Agent Observability
31mBuilding Trustworthy AI: Open Source Monitoring in Action
16mWhat Did the AI Just Do?: Making LLMs Observable with Open Telemetry
01h 02mObserving AI Applications with OpenLit and OpenTelemetry
40mKeep AI on Track: Guardrails + OpenTelemetry Observability
30mFrom Black Box To Insight: Observability for AI Agents in Production
29mNon-Invasive AI Agent Observability With OBI