Méthode Aristote · 2026
Design system & design-to-code tooling for an AI EdTech
As Méthode Aristote’s first Design Engineer, I own the design and front-end layer of a tutoring platform that blends human tutors with generative AI.
The context
Méthode Aristote is reinventing academic tutoring, from middle school to the final year of high school, by combining human guidance with generative AI. I joined as the first Design Engineer of a tight product-and-engineering team where everyone designs, codes, and ships.
The interface ran on stock shadcn at the time: functional, but generic. Nothing set Méthode Aristote apart from yet another dashboard. My brief has two moves: give the platform a coherent identity, and smooth the handoff from design to code for the whole team.
What I own
I hold the design and front-end layer, and I design alongside the rest of the team rather than upstream of it. Day to day, that spans four fronts.
UI & Design System
I built a v2 Design System, still shadcn-based but aligned with Méthode Aristote’s visual identity. Its tokens are synced with the codebase, so decisions made on the design side land on the development side with no friction and no re-transcription.
Bridging design and code
To get design and code speaking the same language, I set up the tooling that ties them together: Figma MCP, a design-tokens pipeline, a Storybook as the components’ source of truth, and visual regression tests with Argos to catch drift before it reaches production.
Designing and prototyping
I design and prototype features wherever it’s fastest in the moment: in Figma, with Claude Design, or straight in code. That ranges from tutor-facing surfaces to dashboards and student-side activities.
Tooling the team with AI
Beyond the product, I tool the team: skills, internal agents, and workflows that speed up our build. The most structural one is an LLM agent wired to the Design System: it knows our components and tokens, and generates coherent interfaces on the first pass.
Impact
The mockups grew calmer. They shed the tells of generative AI and now hold to the Design System. More importantly, screens became deliverable by every profile on the team, not just designers: a PM or QA can assemble a sound interface from reliable building blocks.