⚡ Building AI QoE Analysts: Learnings from Data Agents and MCP Servers

What happens when you give AI agents access to video streaming metrics? We set out to explore this using Mux's recently released MCP server, and discovered that building effective AI analysts requires rethinking how we present data and manage context to LLMs.

This lightning talk shares practical lessons from our exploration of AI-powered QoE analysis, including:

• Context Engineering Patterns: How we're learning to structure metrics, baselines, and domain knowledge for AI reasoning

• Agent Orchestration Insights: Our experiments with chaining specialized agents for different analytical tasks

• Data Presentation Strategies: Evaluations into presenting video performance data in ways that enable meaningful AI analysis

• Prompt Architecture Evolution: The journey from simple queries to domain-aware interactions

Adam shares insights from our ongoing experiments, unexpected challenges he's encountered, and the evolving patterns being discovered for making AI agents effective partners in understanding video streaming performance.