Abirami Sadasivam

Architecting a multi-agent AI orchestrator on LangGraph — parallel dispatch, conditional routing, and typed state reducers coordinating concurrent domain agents and config-driven experts that autonomously triage production SWAT incidents, cutting mean time to resolution from over an hour to a matter of minutes. Alongside it, an MCP server on FastAPI that exposes automated health checks across our data sources as AI-consumable tools with label-based discovery and runtime tool selection, backed by a self-learning memory layer on PostgreSQL and pgvector for semantic pattern retrieval across investigations.

Designed and built full-stack workflow orchestration engines with Spring Boot and Angular, automating network device OS upgrades across multi-vendor topologies and replacing high-risk manual runbooks with validated, rollback-capable pipelines that saved hundreds of hours of operational effort a year. I also developed an interactive infrastructure topology visualization tool on Cytoscape.js with real-time node and edge mapping and integrated health diagnostics, and built autonomous self-healing systems that detect and remediate server and host alerts in real time without human intervention.

Worked on backend services and data infrastructure, building REST APIs with Spring Boot and engineering scalable Airflow DAGs aligned with business workflows. Led an end-to-end data migration effort from MySQL to PostgreSQL, modernizing the data layer while setting up pipeline monitoring and large-scale analytics.

Developed and maintained automated test suites using Selenium and Cucumber, streamlining regression testing workflows and improving structured bug reporting through Azure DevOps.