Ruben Saucedo...

I'm Ruben, a Software Engineer building production full-stack TypeScript systems and applied AI agent experiences. I work across React and TypeScript interfaces, server-side agent orchestration, data grounding, MCP integrations, model evaluation and selection, observability, accessibility, and web performance.

At Microsoft, I own product-specific skill and synthesis architecture for two conversational-agent personas, spanning intent routing, MCP tools, business-data grounding, response synthesis, and graceful degradation when individual skills fail. I also delivered an agent-configuration experience for defining behavior, managing knowledge sources and access boundaries, testing responses, and publishing agents for live use.

I reduced P95 latency for an end-to-end agent response pipeline by approximately 75–82% through multi-source context optimization, chunking, parallel retrieval and model execution, model benchmarking, and telemetry. I also tech-led a Microsoft 365 administration redesign rated 4.1 stars from more than 4,000 marketplace ratings, halving time to interactive while modernizing theming and interface patterns.

Before Microsoft, I built a React lending dashboard and an Angular conversational interface at Wizeline, then worked on enterprise network and ASA firewall issues at Cisco Systems. My personal and open-source projects include KetzalFit (opens in a new tab), lectoria, and kai.