Estevan Seyfried

I’m a Software engineer with 4+ years at Microsoft, where I build platform-layer systems and “own it end to end” on ambiguous technical problems. My focus spans AI agent tooling—especially Model Context Protocol (MCP)—and secure platform infrastructure that supports large-organization workflows.

In my current role, I build identity-protected reporting by aggregating security telemetry from Kusto into Entra-based views. I also drive automation across a ~30-control operations program by assessing controls for automation opportunity, scoring impact, and piloting the highest-leverage targets.

Previously, as part of the AI Agent Platform team, I helped create an OS-level platform for managing MCP-based AI agents and tools. I contributed to server lifecycle management, client identity, runtime behavior, and the tool/agent registry, and I owned OS-level registry integration end to end—covering client enumeration APIs, tool/server registration, consent-flow plumbing, and the test packaging needed to ship reliably.

I’ve also designed and shipped public SDK APIs for multi-window management and accessibility, and I prototyped AI integration scenarios (screen-region detection and spreadsheet action automation) that became reusable integration guides and test apps for downstream teams. I’m eager to transition into empirical AI safety and security research while continuing to build secure, dependable AI systems.