Foundations
Shared rules that give the interface consistent rhythm, contrast and behavior.
CRIS/OS · SYSTEM + AI LAB
Design decisions turned into observable architecture.
This lab separates three topics that competed for attention on the Home page: the Design System, the multi-framework strategy and the evolution toward AI-assisted products.
The Artificial Intelligence specialization has been in progress since August 2026. The following lines are explorations and roadmap items, not completed professional experience.
← Back to portfolio01 · DESIGN SYSTEM
A shared language between intent and execution.
CRIS/OS works as product infrastructure: it defines reusable decisions, reduces ambiguity and connects design, content, accessibility and code.
Tokens
Variables that express intent and allow themes to change without coupling components to visual values.
Components
Documented UI contracts covering purpose, API, content, keyboard and edge cases.
--color-action-primary: var(--primitive-green-300);--color-surface-raised: var(--primitive-neutral-900);--color-border-focus: var(--primitive-yellow-300);Implemented in this prototype
02 · FRAMEWORK STRATEGY
Three technologies, three clear responsibilities.
Versatility is better demonstrated through sound decisions than by loading three frameworks into one view. Each demo will be isolated, measurable and removable.
01IMPLEMENTED
Vue / Nuxt
Editorial shellBilingual content, SEO, navigation, Design System and static generation.
Current production02ROADMAP
React
Agentic experimentConversational UI, sources, traces and response evaluation.
/lab/react03ROADMAP
Angular
Enterprise flowRules, forms, tables, validations, states and approvals.
/lab/angular03 · AI SPECIALIZATION PATH
Learn, prototype, evaluate and document.
The specialization will progressively become public evidence through small experiments, reproducible cases and responsible evaluation.
AI Product Design
Expectations, human control, explainability, trust and error recovery.
Next deliverable: UX heuristics for Ask Cris.Agents + RAG
Structured sources, cited retrieval, tool boundaries and answers that acknowledge uncertainty.
Current base: local fallback grounded in public sources.NLP + automation
Intent classification, responsible extraction and automation of repeatable tasks.
TODO: define exercise, dataset and evaluation criteria.Computer vision
Exploring visual accessibility, classification and graphic asset analysis.
TODO: scope a case without sensitive data.AI evaluation
Test sets, source precision, abstention, safety, cost and user experience.
Next deliverable: an eval matrix for Ask Cris.Responsible AI
Privacy, data minimization, transparency, human oversight and risk analysis.
Guardrails documented before integrating a real API.Nothing is presented as a professional result until evidence, context, tests and publication approval exist.