ProductOS
Own product · Open source · 2025 · Product Design · AI · Full-stack
The whole product cycle in one place, with an AI Mentor that takes part in the decisions. I treated my own problem as a product and shipped it to production.
The problem I lived: The whole product cycle, scattered across loose tools.
I was a designer on real projects with no consolidated view of research, roadmap, metrics and tests. Each front lived somewhere else: a spreadsheet here, Notion there, lost docs. No tool brought the product lifecycle together from the point of view of the person designing it.
The product insight: Treat my own problem as a real product.
It wasn't "I'll learn React". It was "I'll solve this in a structured way": I defined the user (me), mapped the jobs-to-be-done, prioritized features and picked the stack. Moving from task to product is what holds up everything that came after.
What I built: A modular dashboard, from discovery to handoff.
Six phases, each with its own tools: discovery, definition, ideation, design, validation and handoff. At the center, an AI Mentor with tool calling that reads and writes structured data in the database. It knows which personas exist, what has been decided and the audit history, with no copy-pasted context.
Real use at Klini: Not a static portfolio piece. It runs on a real case.
ProductOS is being used in the redesign of Klini Saúde's portals: UX audits, design system documentation, component management and visual bug tracking. It is the product proving itself on a real use case, in production, not in a demo.
The modules, by phase
- Discovery: Interviews, VoC, review and meeting analysis, benchmark
- Definition: Personas, JTBDs, HMW questions, strategic context
- Ideation: Feature backlog, user stories, OKRs, hypothesis map
- Design: Figma frame import, heuristic and WCAG audits, AI redesigns
- Validation: Usability tests, heatmaps, product metrics
- Handoff: Technical docs, release notes, sprint retro
Open for anyone to look inside.
ProductOS is open source. The code, the architecture and the decisions are on GitHub.