Microsoft

Sofware Engineer

2024/09 - Present

Responsibilities & Achievements

• Built an On-Call Engineer AI Agent (FastAPI, LangGraph, MCP) that autonomously triages incidents across Microsoft's vast incident database — using search ranking algorithms to filter thousands of incidents to a relevant few, then leveraging an LLM for root-cause analysis, mitigation steps, and team routing, with a queue-based integration (Azure Service Bus) to broadcast AI-generated insights to downstream engineering platforms.
• Productionized an embedding-based incident similarity reranker with data scientists, transforming proof-of-concept notebooks into a resilient Python package with autonomous retraining; owned reliability of the team's primary ML endpoint, traced cascading peak-hour failures, and drove a 65% daily cost reduction with a 50% incident drop.
• Became an internal Microsoft maintainer of AutoGen, Microsoft's multi-agent framework, contributing message-handling logic for multi-agent orchestration and bug fixes for the 0.4 release.