2026 · Under NDA · Details are kept general and no screens are shown
2026 · Under NDA · Details are kept general and no screens are shown
Simplifying a high-friction flow
Simplifying a high-friction flow
Simplifying a high-friction flow
Redesigned the core experience of a B2C product for a German startup - fewer steps, consistent responsive UI, investor-ready in 3 weeks.
Redesigned the core experience of a B2C product for a German startup - fewer steps, consistent responsive UI, investor-ready in 3 weeks.
Redesigned the core experience of a B2C product for a German startup - fewer steps, consistent responsive UI, investor-ready in 3 weeks.
Product
Product
B2C (under NDA)
B2C (under NDA)
Role
Role
Sole Product Designer
Team
Team
Founder, CTO, Investor (input)
Timeline
Timeline
May–Jul 2026
CONSTRAINT
CONSTRAINT
70+ existing screens, no design file, no design system, investor deadline
Outcome
Outcome
Corrected user flow · Reduced cognitive load · Fewer steps · Consistent responsive UI · Delivered within 3-week window
Corrected user flow · Reduced cognitive load · Fewer steps · Consistent responsive UI · Delivered within 3-week window
Corrected user flow · Reduced cognitive load · Fewer steps · Consistent responsive UI · Delivered within 3-week window
Product
heyroom; a C2C roommate matching app basis vibes, 30k users, Germany
Role
Sole Product Designer
Team
Founder, CTO, Developer
Timeline
Oct–Dec 2024
CONSTRAINT
One developer, different time zone, investor deadline in 8 weeks
WEBSITE
The situation
The situation
The situation
A product with 70 screens and no source file.
A product with 70 screens and no source file.
No design file
70+ screens, no Figma source, no single source of truth for the UI.
Room search ends once you move in. A monthly subscription charges users for time they don’t need.
No design system
No reusable components, no spacing rules, no documented patterns.
No reusable components, no spacing rules, no documented patterns.
3-week deadline
Investor demo deadline. One developer. No time for a full rebuild.
Investor demo deadline. One developer. No time for a full rebuild.
Key decisions
Key decisions
Key decisions
Extracting what existed instead of starting over
Extracting what existed instead of starting over
The team's first instinct was to design everything from scratch. I pushed back. Building a new design system for 70+ screens would take weeks the timeline didn't have. Instead, I used Claude Code to extract existing UI elements directly from the codebase - colours, spacing, typography, component patterns and cross-checked each one with the developer to confirm accuracy. From these elements, I built prototype versions to review with the founder, investor, and developer. The result: a working foundation in days, not weeks.
The team's first instinct was to design everything from scratch. I pushed back. Building a new design system for 70+ screens would take weeks the timeline didn't have. Instead, I used Claude Code to extract existing UI elements directly from the codebase - colours, spacing, typography, component patterns and cross-checked each one with the developer to confirm accuracy. From these elements, I built prototype versions to review with the founder, investor, and developer. The result: a working foundation in days, not weeks.
Redesign the journey, not individual screens
Redesign the journey, not individual screens
Rather than polishing screens one at a time, I mapped the primary flow end-to-end and looked at it as one continuous experience. The core problem became clear: too many steps & CTA's, inconsistent structure between them, and confusing feedback after the key action with slow animations. I reorganised the flow around the user's actual goal, that is, understand the choice, act, get confirmation, and removed everything between those moments including irrelevant animations.
Rather than polishing screens one at a time, I mapped the primary flow end-to-end and looked at it as one continuous experience. The core problem became clear: too many steps & CTA's, inconsistent structure between them, and confusing feedback after the key action with slow animations. I reorganised the flow around the user's actual goal, that is, understand the choice, act, get confirmation, and removed everything between those moments including irrelevant animations.
Responsive from the start
Responsive from the start
The product needed to work across three breakpoints. Instead of designing desktop first and adapting later, I treated responsive behaviour as part of each design decision from the beginning - how hierarchy, spacing, and controls should shift across screen sizes while keeping the interaction model the same.
Revenue was the obvious KPI, but I pushed back. With limited room supply, low revenue wouldn’t mean the design failed. It could just mean there weren’t enough rooms to apply to. The metric that would actually tell us if the model & design was working was engagement: were users coming back daily to use their 3 free applications? Were drop-offs increasing or staying flat? We tracked both.
The goal: Use AI to extract what exists. Redesign the journey. Ship within the constraints.
Design Evolution
Design Evolution
Design Evolution
01
01
01
Claude Code for prototyping
Claude Code for prototyping
With three weeks and one developer, long design-review loops would have killed the timeline. I used Figma for exploration and visual direction, then Claude Code to rapidly prototype parts of the experience. Instead of waiting for a polished Figma flow, I could discuss an idea with the founder, prototype it, demonstrate the interaction, and get feedback in the same conversation. Stakeholders reacted to working behaviour, not static screens.
With three weeks and one developer, long design-review loops would have killed the timeline. I used Figma for exploration and visual direction, then Claude Code to rapidly prototype parts of the experience. Instead of waiting for a polished Figma flow, I could discuss an idea with the founder, prototype it, demonstrate the interaction, and get feedback in the same conversation. Stakeholders reacted to working behaviour, not static screens.
02
02
02
Claude Code for handoff
Claude Code for handoff
A pixel-perfect Figma handoff wasn't the highest-value use of the remaining time. Instead, I delivered annotated specs, responsive behaviour guidance, reusable patterns, and working prototypes. I also used Claude Code to generate front-end component references aligned with the team's existing tech stack not as a substitute for engineering, but as an implementation reference that reduced ambiguity during handoff.
A pixel-perfect Figma handoff wasn't the highest-value use of the remaining time. Instead, I delivered annotated specs, responsive behaviour guidance, reusable patterns, and working prototypes. I also used Claude Code to generate front-end component references aligned with the team's existing tech stack not as a substitute for engineering, but as an implementation reference that reduced ambiguity during handoff.
Outcome
Outcome
Outcome
Corrected user flow · Reduced cognitive load · Fewer steps · Consistent responsive UI · Delivered within 3-week window
The redesigned experience was implemented within the three-week window and is currently in use.
The redesigned experience was implemented within the three-week window and is currently in use.
Reflections
Reflections
Reflections
Extracting before rebuilding
Extracting before rebuilding
The team assumed starting from scratch was the only option. Using AI to pull elements from the existing codebase gave me an accurate foundation faster than any manual rebuild and it matched what was already in production.
Working directly with a founder means priorities shift daily. Early on, last-minute changes were derailing both my timelines and the developer’s. I started documenting every decision and the reasoning behind it which made it easier to push back, and easier for the founder to see when a change had downstream costs.
Handoff is defined by what the team can build from
Handoff is defined by what the team can build from
A polished Figma file that can't be acted on within the deadline isn't a good handoff. Matching the deliverable to the constraint mattered more than completeness.
I’d design for German copy from the start. Retrofitting the multilingual layout cost 3–4 days we didn’t have.