Bloomp — from style guide to a scalable multi-tenant design system.

A product that existed mostly on paper.

When I joined Listra, Bloomp was already under construction, but without components or a design system to make it real.

There was a simplified style guide in Figma. It wasn't componentized. It wasn't scalable. It had no real adaptation to the product's white-label model, and had serious accessibility issues in both light and dark modes. No design system, no structured discovery, no architecture built to grow.

As the only designer on the team product, my job was to turn that into a foundation before the first enterprise client came in.

Three problems, one timeline.

Ask before you draw.

I started with structured discovery. Used NotebookLM for desk research on competitors, market patterns and the first client's context. In parallel, I built personas based on real product user profiles, JTBD mapping what each persona needs to accomplish — and why, journeys for both the app and the admin identifying friction and opportunities, and flows prioritized by usage frequency and business impact.

I also built a white-label readiness map — an audit of everything that needed to become a variable to support tenants — and gamification explorations aligned with JTBD.

All documented across four FigJam boards, shared with the team as the single source of truth.

Foundation first.

The DS isn't a library of pretty components. It's the contract between design, engineering and every future tenant.

App mobile + admin panel, redesigned in parallel.

With the foundation in place, I redesigned both products. On the mobile app, I fully restructured the information architecture and simplified critical engagement flows. On the admin panel, I worked in coherence with the app — same language, same patterns — so the enterprise client has predictability when managing content, users and experiences.

Highlights: complete IA restructuring, new gamification system tied to discovery JTBDs, revised empty/error/loading/success states across every critical flow, and all flows adapted for multi-tenant without duplication.

AI as a lever, not decoration.

I used AI as a real productivity lever across three moments of the process.

One designer, eight engineers, zero PDFs.

Being the sole designer on a full engineering team forced a process that's both tight and scalable:

• Live specs in Figma — no PDFs. Components, tokens and states organized so engineering consumes directly.
• Documented decisions — every design choice has a "why" captured in FigJam or Notion, so it's traceable.
• Presence in refinement — I sit in ceremonies to catch technical blockers early and adjust design when it makes sense.
• Direct pairing — quick loops with tech lead and devs on complex interactions before the ticket is written.

What's shipped, what I learned, what's next.