Multi-Agent AI Operations System for a Confidential Enterprise
Stateful multi-agent orchestration for enterprise IT operations, coordinating database NL-to-SQL, IT service management automation, and infrastructure monitoring workflows.
Business Challenge
Enterprise IT operations involved recurring manual work across database access, service management, and infrastructure monitoring. The project aimed to reduce manual coordination by providing one stateful AI workflow that could route operational requests to specialised agents while maintaining context across multi-step tasks.
Solution Delivered
As an AI Engineer at FPT Software, I architected and implemented a stateful LangGraph orchestration layer coordinating three specialised agents: a database NL-to-SQL agent, an IT service management automation agent, and an infrastructure monitoring agent. The orchestration layer managed workflow state and coordination across multi-step requests. The solution was packaged with Docker and deployed through a CI/CD pipeline. My contribution focused on orchestration design, agent workflow integration, and deployment through the CI/CD workflow.
Outcomes Achieved
The pilot reduced manual IT operations effort by approximately 40%. It brought three specialised workflows behind one orchestration layer and reduced the manual coordination required across database, service-management, and infrastructure-monitoring tasks. This is a pilot result and should not be interpreted as a guaranteed organisation-wide saving.
Measurable Business Outcome
Reduced manual IT operations effort by approximately 40% during the pilot phase.