Microsoft
Senior Software Engineer
MCIS AI Innovation · Previously Azure Core Insights
- Designed, implemented, and scaled the Model Hosting Framework (MHF) end to end: a Kubernetes platform serving 40+ production endpoints through a FastAPI gateway for authentication, routing, and streaming, supporting nearly 1M calls per month at 99% availability.
- Built incident-routing automation with 4 agents covering 66 of 69 CIS queues and approximately 292.8K incidents per year, reducing average time-to-engage by 58% and routing hops by 72%.
- Shipped DERA and the CARET front end to summarize and root-cause 30K+ support tickets per day for 50+ Azure service PMs; scaled serving across 30+ Azure OpenAI model endpoints.
- Led the Agency Review rollout for a 100-repository scope, with adoption monitoring and review-evaluation guardrails aimed at consistent feedback and shorter review cycles.
- Built Tortoni, a scalable AKS-based load balancer across 30 OpenAI endpoints in multiple regions, and published a Python package.
- Launched CIS Incident MCP in production and AME with 5 incident-intelligence tools for routing, summaries, classification, similarity search, and ownership; enabled Incident Investigator with end-to-end validation and usage telemetry for ~120 daily active users. Incident-intelligence workflows average ~500 tool calls/day, with a goal of reducing incident time-to-mitigate.
- Serve as Service Security Champion for MCIS AI Innovation; previously completed and drove 500+ security remediations across Azure Core Insights SFI Waves 2–4, helping raise Get Secure Score above 650 across services.
- Led SFI Waves 4–6 with ~400 security action items during May–October 2025; introduced MHF container-hygiene guardrails that reduced vulnerability findings from 3,000+ to <50.
- Patent contributor for a Retrieval-Augmented Generation platform blending vector stores, identity enrichment, and agentic code generation to deliver real-time, context-rich answers and executable workflows.
- Enhanced Wayfinder support recommendations with +3 coverage, +1 accuracy, and 60% latency reduction while modernizing CI/CD for Wayfinder and Service Predictor (12% maintenance cost savings).
- Shipped Azure Case Submission Autosuggest, Azure Workload Identification (clustering + BERT topic modeling), and an IcM Team Prediction service powered by Azure Cognitive Search running on AKS.