Ruben Saucedo...
Senior fullstack engineer building performant AI agent platforms.
Ketzal — interactive pet
About me
For the past 16 months at Microsoft I've been building an AI agent platform — orchestration, agent protocols, eval harnesses, and the production telemetry that keeps multi-step reasoning honest at enterprise scale. The work is grounding agents against governed customer data: latency budgets, retries, and UX that respects users when models misbehave.
The fullstack foundation is what makes that possible. Over six years at Microsoft I've owned complex frontend surfaces end-to-end — a 30–40% performance gain on a product used by 100M+ people every day, a full redesign that shipped as the standard UI (4.1★ on the public marketplace), and serving as my team's Performance and Accessibility Champion. The common thread: making things fast, accessible, and resilient at a scale where small wins compound.
I'm Ruben. Six years at Microsoft, joined in 2020. Before Microsoft I shipped production React at Wizeline for international clients, and I started my career at Cisco Systems as a Network Security Engineer. UPIITA in Mexico City, with an exchange in the Basque Country.
On the side I run KetzalCode, an AI trainer platform I architected from scratch. In 2018 I co-invented an on-device machine learning system for Mexican Sign Language translation — it's a national patent.
Now
updated: 2026-05-15
This week: tightening reliability on an AI agent platform — latency budgets, retries, telemetry.
Shipping a platform that grounds AI agents against governed customer data. Owning the fundamentals — performance, telemetry, resilience — that make agentic products reliable in production. Mentoring teammates and authoring design specs.
On the side, building features into KetzalCode.
Currently mastering agentic architecture — eval harnesses, tool routing, and the failure modes that only surface in production.
Contact
- LinkedIn:
- linkedin.com/in/enrique-saucedo-diaz
Open to connect. Reach out via LinkedIn.