At Microsoft Azure Storage, I delivered proactive detection of failing power supply units across 250K+ servers, helping lift fleet PSU health from 96% to 99% since July 2025.
I built Proforge to diagnose nodes stuck out of production and run repair recipes, triggering recovery on 6K+ nodes and restoring about 37% and 955 PB of capacity since August 2026. I also developed RackOffload to safely drain and restore racks during maintenance, saving an estimated 4,080 manual hours per quarter.
I developed an AI Node RCA agent in Python that brings seven hardware and repair log sources into one timeline, cutting RCA preparation from 2–3 hours to about 3.5 minutes with human review before release. Earlier at Microsoft, I troubleshot Azure customer issues across several services and supported cloud migrations as the Azure Messaging point of contact.
