2025 · UX Lift
2025 · UX Lift
An AI-Native SaaS product I built e2e
An AI-Native SaaS product I built e2e
An AI-Native SaaS product I built e2e
Identified a gap in Berlin's design job market, validated it with 30+ designers, and shipped a working web app in 15 days.
Identified a gap in Berlin's design job market, validated it with 30+ designers, and shipped a working web app in 15 days.
Identified a gap in Berlin's design job market, validated it with 30+ designers, and shipped a working web app in 15 days.
Product
Product
UX Lift, an AI native B2C SaaS product that gives junior designers real use cases to build portfolio work from and send custom made proposals to relevant startups.
UX Lift, an AI native B2C SaaS product that gives junior designers real use cases to build portfolio work from and send custom made proposals to relevant startups.
Role
Role
Founding Product Designer
Team
Team
Solo (product strategy to production)
Timeline
Timeline
Apr–Jun 2025
CONSTRAINT
CONSTRAINT
Limited budget for backend implementation
WEB LINK
WEB LINK
Outcome
Outcome
30+ designer interviews · 100+ early subscribers · Shipped in 15 days
30+ designer interviews · 100+ early subscribers · Shipped in 15 days
30+ designer interviews · 100+ early subscribers · Shipped in 15 days
Product
UX Lift, an AI native B2C SaaS product that gives junior designers real use cases to build portfolio work from and send custom made proposals to relevant startups.
Role
Founding Product Designer
Team
Solo (product strategy to production)
Timeline
Apr–Jun 2025
CONSTRAINT
Limited budget for backend implementation
WEBSITE
The situation
The situation
The situation
I was attending design community events in Berlin and meeting designers at every stage of their career. A pattern kept coming up in conversations, in job boards, & in my own job applications.
I was attending design community events in Berlin and meeting designers at every stage of their career. A pattern kept coming up in conversations, in job boards, & in my own job applications.
Around 90% of product design roles in Berlin required 5+ years of experience. Junior designers were stuck in a loop: they couldn't get hired without shipped work, and couldn't get shipped work without being hired. Bootcamp projects and unsolicited redesigns weren't cutting it. Hiring managers wanted to see real problems solved for real users.
Around 90% of product design roles in Berlin required 5+ years of experience. Junior designers were stuck in a loop: they couldn't get hired without shipped work, and couldn't get shipped work without being hired. Bootcamp projects and unsolicited redesigns weren't cutting it. Hiring managers wanted to see real problems solved for real users.
So I decided to design a product which can solve this problem for junior designers.
So I decided to design a product which can solve this problem for junior designers.
Understanding the problem
Key decisions
Key decisions
Key decisions
Find credible problems
Designers could find fictional prompts and redesign exercises, but nothing grounded in real user behaviour and business context.
Know what makes work strong
After finishing a project, designers weren't sure if their work demonstrated the product thinking hiring teams look for.
Connect work to opportunities
Creating another case study had limited value if designers still had to search separately for relevant companies.
Validate before building
Validate before building
I spoke with over 30+ junior designers and some mid & senior- level too to understand whether this problem was widespread enough to build for. Three needs kept surfacing:
I spoke with over 30+ junior designers and some mid & senior- level too to understand whether this problem was widespread enough to build for. Three needs kept surfacing:
Find credible problems
Designers could find fictional prompts and redesign exercises, but nothing grounded in real user behaviour and business context.
Designers could find fictional prompts and redesign exercises, but nothing grounded in real user behaviour and business context.
Know what makes work strong
After finishing a project, designers weren't sure if their work demonstrated the product thinking hiring teams look for.
After finishing a project, designers weren't sure if their work demonstrated the product thinking hiring teams look for.
Connect work to opportunities
Creating another case study had limited value if designers still had to search separately for relevant companies.
Creating another case study had limited value if designers still had to search separately for relevant companies.
Understanding Product Strategy
Understanding Product Strategy
I attended a product management workshop (Hello PM) specifically to pressure-test the product-market fit question and potential value proposition before committing to a build.
I attended a product management workshop (Hello PM) specifically to pressure-test the product-market fit question and potential value proposition before committing to a build.
Slide from PM Workshop
I analysed these questions before I started the project.
I analysed these questions before I started the project.
Are you the right person to solve this?
How big is the market?
How acute is this problem?
Did this recently become necessary?
Do you have competition?
Do users want it or not?
Idea you want to work for years
Are there good proxies for the business?
Is this scalable business?
Are you the right person to solve this?
How big is the market?
How acute is this problem?
Did this recently become necessary?
Do you have competition?
Do users want it or not?
Idea you want to work for years
Are there good proxies for the business?
Is this scalable business?
Identifying the core problem
Identifying the core problem
The JBTD research surfaced a clear job map: designers needed help across the entire arc that is, from setting goals to skill-building, to finding real projects, to networking, to tracking applications, to staying motivated.
The JBTD research surfaced a clear job map: designers needed help across the entire arc that is, from setting goals to skill-building, to finding real projects, to networking, to tracking applications, to staying motivated.
Trying to solve all of them would have produced a broad but shallow product.
Trying to solve all of them would have produced a broad but shallow product.
So I looked for the point where three conditions overlapped:
So I looked for the point where three conditions overlapped:
High user frustration + weak existing solutions + a meaningful opportunity for AI
High user frustration + weak existing solutions + a meaningful opportunity for AI
Understanding the user journey using JBTD
That led me to focus the first version on:
That led me to focus the first version on:
Finding a relevant product problem → turning it into a realistic design challenge → building proof of skill → connecting that skill to relevant companies.
Understanding Target market
Understanding Target market
Target User
Build with AI to move at founder speed
Build with AI to move at founder speed
I chose Lovable to design and build the complete app from scratch because it let me make product decisions at the speed of my research - Vibe coding, API integrations via Supabase, full responsive app. 15 days from first commit to working product.
I chose Lovable to design and build the complete app from scratch because it let me make product decisions at the speed of my research - Vibe coding, API integrations via Supabase, full responsive app. 15 days from first commit to working product.
Design evolution
Design evolution
Design evolution
01
01
01
Real problems, AI simulated environments
Real problems, AI simulated environments
This was the most important product decision. If UX Lift asked an LLM to invent a generic design challenge, the exercise might look realistic but would still be fictional. That would undermine the core promise so I designed around two separate layers.
This was the most important product decision. If UX Lift asked an LLM to invent a generic design challenge, the exercise might look realistic but would still be fictional. That would undermine the core promise so I designed around two separate layers.
Real-world signal: The starting problem comes from actual UX complaints and pain points. The product uses Reddit's public search API to find real user frustrations associated with a specific industry or product category.
Real-world signal: The starting problem comes from actual UX complaints and pain points. The product uses Reddit's public search API to find real user frustrations associated with a specific industry or product category.
Simulated context: AI then turns that signal into a structured design challenge by adding realistic but explicitly simulated elements that is - a business objective, product constraints, target users, design scope, and suggested KPIs.
Simulated context: AI then turns that signal into a structured design challenge by adding realistic but explicitly simulated elements that is - a business objective, product constraints, target users, design scope, and suggested KPIs.
02
02
02
Designing the core experience
Designing the core experience
UX Lift uses AI at multiple points in the experience, but I didn't want the product to feel like one large chatbot. Instead, I designed AI to take a different role depending on what the user was trying to accomplish.
UX Lift uses AI at multiple points in the experience, but I didn't want the product to feel like one large chatbot. Instead, I designed AI to take a different role depending on what the user was trying to accomplish.
CLICK EACH CARD TO VIEW STEPS
Step 1: What kind of product? (B2B / B2C)
Step 2: Which industry? (Fintech · HealthTech · E-commerce · etc.)
Choosing the industry
Step 3: Choose a problem (UX Lift surfaces relevant problems in that space)
Step 4: Generate the brief (AI transforms the problem into a structured challenge)
Step 5: Case study review by AI
Step 6: Case study review by Experts
Step 7: Matching relevant Startups
03
03
03
Edge Cases & technicalities
Edge Cases & technicalities
Because much of UX Lift relies on AI-generated outputs, I designed for situations where the system might return incomplete or unusable results.
Because much of UX Lift relies on AI-generated outputs, I designed for situations where the system might return incomplete or unusable results.
No relevant Reddit(API) insights: I defined a fallback experience for cases where Reddit couldn’t surface a relevant user problem, instead of leaving the user with an empty or misleading result.
No relevant Reddit(API) insights: I defined a fallback experience for cases where Reddit couldn’t surface a relevant user problem, instead of leaving the user with an empty or misleading result.
Unsupported file formats: Uploaded files are validated before analysis. If the file type isn’t supported, users receive a clear error message and guidance on accepted formats.
Unsupported file formats: Uploaded files are validated before analysis. If the file type isn’t supported, users receive a clear error message and guidance on accepted formats.
Technical implementation: UX Lift uses Supabase Edge Functions for backend workflows, Firecrawl for web data extraction, and Gemini for AI-powered analysis and generation.
Technical implementation: UX Lift uses Supabase Edge Functions for backend workflows, Firecrawl for web data extraction, and Gemini for AI-powered analysis and generation.
Shipping
Shipping
Shipping
I used Lovable for the full frontend and design implementation, Supabase for authentication and backend logic, the Gemini API (via Lovable AI Gateway) for case study evaluation, and Reddit's public API for pain-point sourcing and Firecrawl for scraping.
I used Lovable for the full frontend and design implementation, Supabase for authentication and backend logic, the Gemini API (via Lovable AI Gateway) for case study evaluation, and Reddit's public API for pain-point sourcing and Firecrawl for scraping.
15 days after starting the build, UX Lift was a working product users could sign up for and try it out.
15 days after starting the build, UX Lift was a working product users could sign up for and try it out.
Outcome
Outcome
Outcome
30+ designer interviews · 100+ early subscribers · Community feedback from Lovable Shipped challenge
I marketed the app over LinkedIn and in Hackathons. It attracted over 100 early subscribers and generated substantial feedback on the Lovable community as well. Designers were responding to the core value proposition that is, real use cases which are connected to real opportunities.
I marketed the app over LinkedIn and in Hackathons. It attracted over 100 early subscribers and generated substantial feedback on the Lovable community as well. Designers were responding to the core value proposition that is, real use cases which are connected to real opportunities.
Marketing on LinkedIn
Real feedback from designers
What happened next:
What happened next:
Because of backend challenges (manual data entry via Supabase, credit system complexity, engineering gaps I identified after consulting with 3 engineers), I didn't continue the product further but with advanced AI tools like Claude code, Codex, etc. implementation seem smoother.
Reflections
Reflections
Reflections
Building a product changed how I design products
Building a product changed how I design products
Running the full loop from research, JBTD, scoping, building, launching, collecting feedback gave me a reference point I use in every project now.
AI tools are a capability multiplier, not a shortcut
AI tools are a capability multiplier, not a shortcut
Lovable and Supabase didn't skip the hard decisions but they compressed the time between a decision and seeing its execution. The 15-day build was possible because the 3 months of research and my own experience before it had already answered the important questions.
Designing for AI is a different skill than designing with AI
Designing for AI is a different skill than designing with AI
Building the product taught me both. Using AI tools to ship fast is a workflow advantage. Designing how AI output reaches users: what to show, what to let them control, how to build trust.