Work with me

Creating, auditing and rescuing design systems: tokens, components, documentation, governance and, most importantly, adoption. I've built design infrastructure for global and fast-moving teams like Telefónica, El Español and Metallicus.

Your design system now has a second audience. I make tokens, components and rules machine-readable, so coding agents build with your real components instead of inventing their own. Read more in Design systems for humans and agents.

Designing products where AI does real work: MCP connectors for Claude and ChatGPT, agent and chat experiences, and AI recommendations people can understand and trust. See Ensinova and Sphera.

Review agents, skills, research and QA pipelines that encode your team's design judgment into reusable systems, so quality doesn't depend on who is in the room. See Shipping AI, not just using it.

End-to-end product design for complex domains, from discovery to shipped UI, and hands-on leadership for design teams that are growing faster than their process. See all projects.

Focused projects: a scoped piece of work with a clear outcome, like a design system audit, an MCP connector or a new product flow.

Advisory: ongoing, part-time support for your design and product team on design systems, AI products and design operations.

What does a design system engagement with you look like?

It usually starts with an audit: what exists in Figma, what exists in code, and where the two have drifted apart. From there we agree on one source of truth for tokens, a component set that matches what actually ships, written rules for when to use what, and a plan to drive adoption across teams.

Can you design an MCP connector or AI agent experience for my product?

Yes. I designed and built a live MCP connector for Ensinova that lets school owners run their business from Claude or ChatGPT, with safety enforced by the server rather than by the prompt. I can help you decide what your product should expose to AI assistants, design the conversational flows, and make sure risky actions stay under control.

How do I make my design system usable by AI coding agents?

Agents only know what is written down. Keep one machine-readable source of truth for tokens, give agents your real components with their actual props and variants instead of lookalikes, and write down the rules a senior designer carries in their head: responsive behaviour, contrast requirements, copy and tone. Then let a governance agent check every change against those rules.

What industries have you worked in?

AI and SaaS, education, telecommunications (Telefónica), digital news and media (El Español, Observador), ecommerce, fintech and Web3 (Metallicus), and developer tools (Bitloops, Clutch).

How do we start?

Send me an email at celio.r.pires@gmail.com with a few lines about your product, your team and what you need. I will reply with a few questions and a suggestion for the next step.