Expedia Builds AI Ops Platform with Deterministic Workflows, Not AI Agents

Expedia Group introduced Service Telemetry Analyzer (STAR), an AI-assisted observability platform that helps engineers investigate production incidents by analyzing service telemetry data and generating structured root cause assessments. The system follows a deterministic workflow rather than using autonomous AI agents: it collects telemetry data, performs analysis with domain-specific prompts, aggregates findings, and produces a report with potential root causes and suggested next steps. STAR is implemented as a FastAPI application integrated with Datadog for metrics and an internal generative AI gateway for LLM access. It uses prompt chaining to execute multiple specialized analyses before generating a consolidated diagnosis. The platform focuses on standardized infrastructure telemetry from Kubernetes-based services and JVM applications, including metrics like request throughput, latency, error rates, CPU/memory usage, container restarts, and Java heap usage. Expedia replaced FastAPI background tasks with an async Celery architecture using Redis as broker and result backend to handle I/O-intensive operations and rate limits. STAR is used for production incident investigation, post-incident analysis, Kubernetes troubleshooting, and JVM memory diagnostics. Future plans include adding service dependency information, Model Context Protocol (MCP) tool integration, and conversational interfaces. Prompt management and evaluation are supported by Langfuse.