AI, Software & Cloud Architect
Jose Quesada
I design AI-native and agentic systems for enterprises — and the software, cloud and security architecture they have to run on.
SWAT Team Manager at SOIN — the team that takes the engagements nobody else can unblock: systems failing in production, architectures that stalled for reasons no schedule can fix, and the AI work the organisation has not built before.
Building enterprise systems since 2015Costa Rica
- Agentic AI
- Software Architecture
- Cloud & Kubernetes
- DevOps & Security
- Technical Leadership
Core expertise
See all seven domains- AI & Agentic Systems
- Designing AI systems that have to survive contact with real enterprise estates — real credentials, real APIs, real failure modes.
- AI-Native Software Engineering
- Changing how software gets built when agents are in the loop — without giving up the review discipline that makes it safe.
- Cloud & Platform Engineering
- The platforms underneath — designed for availability and cost, not for a diagram.
- Software Architecture
- Enterprise systems that have to integrate with what already exists, and keep running while they change.
Selected work
An agent platform for enterprise systems
A platform for building and running AI agents against real enterprise estates — real credentials, real APIs, real failure modes — rather than against a demo.
A platform that generates applications from specifications
AI code generation for enterprise delivery, built around determinism and human review rather than around prompting — with a knowledge repository as the source of truth.
A mobile app that turns speech into structured data
A cross-platform field app where the primary input is a person talking, and an AI service behind it turns what they said into validated, structured records.
Current focus
Enterprise agent architecture
What an agent needs beyond the demo — tool gateways, credential isolation, and a failure model somebody can operate.
MCP and A2A
Following both as enterprise integration surfaces. MCP is settling; A2A is early, and I am watching it rather than betting on it.
Retrieval architecture
Ingestion, chunking and retrieval quality. In production, retrieval fails before the model does.
AI-native engineering
Specification-driven development with agents in the loop — and where human review has to stay non-negotiable.
AI-assisted modernization
Using models to read large legacy estates: dependency mapping, architecture discovery, and specifications a human validates.
Platform and gateway architecture
Running agent workloads on Kubernetes behind an API gateway, with tenant isolation, secret management and mutual TLS as platform concerns rather than per-service afterthoughts.
Featured credentials
View all certifications
Agentic AI Business Solutions Architect
Microsoft
Machine Learning Engineer – Associate
Amazon Web Services (AWS)
Azure AI Engineer Associate
Microsoft
DevOps Engineer – Professional
Amazon Web Services (AWS)
Security – Specialty
Amazon Web Services (AWS)
DevOps Engineer Expert
Microsoft
Power Platform Solution Architect Expert
Microsoft
Developer – Associate
Amazon Web Services (AWS)
Writing
August 30, 2026
From Software Engineering to AI-Native Engineering
AI did not make software engineering faster. It moved the bottleneck from writing code to deciding whether the code is correct — and I found that out by pointing an agent at my own site.
Beyond the work
I live in Costa Rica and I am a solo parent to two kids, nine and seven. Most of what I do outside work involves them — hiking, zip-lining, horseback riding, whichever adventure is next. I play golf and football, read more than I manage to finish, and watch more anime than I will defend in public. I have been learning Japanese since 2019, slowly and stubbornly, and in 2027 the three of us are going to Japan for sakura season. The curiosity that keeps me pulling apart new tools at work is the same one that keeps me at the language.
Beyond the work
