Senior Software Engineer

• Designed and built an event-driven pipeline that enriched third-party enterprise content ingested through connectors, using author and modifier metadata plus content-derived people mentions to route documents from tenant shards to relevant user shards for Copilot grounding; processed 366M references daily into 32.4M document-to-user links and 2.5M personalization signals, with associated content showing 18% engagement and a 12% Copilot citation rate.
• Expanded candidate discovery for enrichment by resolving access-control groups on ingested files into authorized users and routing permitted content to their user shards for Copilot grounding; identified groups with fewer than 20 members as the optimal inclusion threshold, with added content achieving 20% engagement and a 14% Copilot citation rate.
• Designed and shipped an A/B experimentation platform for connector-ingested enterprise content, aligning with architects on a scalable approach that enables teams to measure how relevance signals affect AI-generated answers; supported two successful experiments within its first quarter.
• Reduced end-to-end content-delivery latency from 3 days to an average of 9 hours by tracing items from connector ingestion through identity processing and final delivery to user shards, isolating bottlenecks and prioritizing reliability improvements across the entire pipeline.
• Built a scheduled AI agent that audited incident hygiene across 13 services each week, parsing incident metadata against security and compliance rules, alerting service owners to approaching remediation deadlines, and standardizing data for future automated detection and remediation. This saved 4 hours per dev team per week that was earlier used to triage and fix incident hygiene.

Technologies: .NET, distributed systems, event-driven architecture, experimentation, Microsoft Graph, AI grounding

• Owned end-to-end frontend development of a Microsoft 365 Copilot agent debugging tool serving 30K+ monthly active users, partnering with UI designers to surface agent actions, capability responses, and orchestration failures through Copilot Chat; shipped from initial design through worldwide production release.
• Designed, built, and shipped 12 releases of Microsoft.DeclarativeAgents.Manifest, a public .NET SDK that reads, writes, and validates API plugin and declarative agent JSON manifests; implemented 90% of critical features used by marketplace and Responsible AI systems, with adoption eliminating plugin-manifest authoring errors at runtime.
• Built TypeSpec tooling for Microsoft 365 Copilot agents, enabling developers to describe agents across the Microsoft ecosystem with simplified, type-safe definitions instead of complex JSON manifests; increased developer adoption from 13% to 19% and showcased the tooling at Microsoft Build.
• Built TypeSpec tooling for Microsoft Graph, a library that enables API producers to describe Microsoft Graph APIs in TypeSpec and generate CSDL instead of authoring complex, error-prone CSDL by hand; improved OpenAPI fidelity and reduced weekly SDK-correctness issues from 300 to under 40, with remaining issues limited to services not yet onboarded.
• Implemented critical features for Microsoft Graph Explorer, a developer application serving 400K monthly active users for testing Microsoft Graph APIs; increased test coverage by 70%, introduced Playwright end-to-end tests that caught UI regressions before production, and helped sustain zero Priority 1 bugs for a semester.

Technologies: TypeScript, React, TypeSpec, C#, .NET, OpenAPI, CSDL, Azure, Playwright

• Led development of an IoT monitoring platform across approximately 10 boreholes, building a Java Spring Boot ingestion service for hourly telemetry on aquifer levels and daily water extraction with centralized Grafana visualization; reduced field visits by 75%, from four monitoring visits per month to one monthly battery-replacement visit.

Technologies: Java, Spring Boot, Grafana, IoT, embedded systems, cloud services