Accelerating aircraft IFEC diagnostics with agentic AI on AWS
aws ·
Panasonic Avionics Corporation, a provider of in-flight entertainment and connectivity systems, partnered with AWS to build an agentic AI diagnostic system. The manual process of correlating logs and metrics across thousands of fleet configurations took hours, so the company aimed to reduce MTTD and MTTR. The solution uses a multi-agent workflow with Amazon Bedrock, SageMaker, and AWS Glue, featuring a trend analyzer, parallel diagnostic agents, and an LLM-powered summarizer. Operational data is ingested and normalized into a data lakehouse, and a domain ontology ensures consistent comparison across fleet variants. Diagnostic agents, orchestrated by SageMaker with LangGraph and Strands Agents SDK, analyze correlations, system checks, and log patterns in parallel, reducing investigation time from hours to minutes. The system retrieves past incident resolutions via vector search and generates grounded, severity-categorized reports with human oversight. As a result, Panasonic achieved 20-40% improvements in operational efficiency and reduced repetitive investigative work.
aws