Jaya Advani.

A homeowner with a leaking sink wants it fixed, not a list of strangers to vet. So I designed a product that finds the one pro they can trust.

At FixerUp, an AI-powered home repair marketplace, I was the sole designer across the user app and the website (desktop and mobile web). I took it from an ambiguous concept to a shipped, high-fidelity beta in about 1 month, and the core of the work was one decision i.e to replace the usual browse-and-compare marketplace with a single, trustworthy AI match.

Jump to the Main Screens or Design System.

Onboarding Screen_Start_screen_option2.png

Onboarding Screen_Category_Selection_Screen.png

THE BET

Every home-services marketplace works the same way, search, browse pros, compare profiles, read reviews, message a few, wait for quotes, decide. That is a shopping task, and it lands on someone at the exact moment they have the least patience for one, because something is broken and they are stressed.

FixerUp's bet was to remove the list. The AI does the vetting and presents one matched pro. My whole job became earning, inside that single match, the trust a list normally asks the user to build for themselves.

PROJECT SPECS

Role: UI/UX Designer (sole designer)
Duration: April 2026 - Present
Platform: Mobile (iOS, Android) and dual-sided marketing web
Team: 3 Founders (CEO, COO, CTO), Engineering team
Tools: Figma, Claude Design, Claude, Adobe Express, CapCut, RunwayML, Krea AI, Notion

NUMBER STORIES

3 Mins

Instant quotes

The average time the AI takes to turn a described problem into an avg. quote.

65

Different Design States

Designed for the project and quote elements, so the app always shows where a job stands.

~1 month

Zero to beta

From an ambiguous AI built concept to a shipped, high-fidelity beta, as the sole designer.

03

Tiers for each User

Service tiers designed for the three user segments the research uncovered.

1. One match, not a list.
A stressed user in an urgent moment doesn't want options, they want a resolution. So I removed the search results, the filters, and the comparison UI entirely. The cost: every gram of trust a list builds through choice now had to live in the match itself, through visible reasoning, verification, and fit.

2. Three quotes, not one, won with data.
The founders wanted to keep it simple: one price for everyone. I pushed back, because the research showed our users weren't one group, they split into the three segments above, with different budgets and needs. I brought the segmentation, not an opinion, and the team will now be shipping a three-tier model: three quotes and three service levels for three user groups. This is also why the journey forks at stage 4, the same decision shows up as both better economics and a better-tailored path.

3. An AI you can see and override.
People trust what they understand. The match always shows its reasoning, and I designed deliberate empty, loading, and error states so it never feels like a black box. When the AI isn't confident enough to price a job, it doesn't bluff a number, it offers a free inspection so a fixer can assess in person first. And if the match isn't right, the user keeps control two ways: chat directly with the matched fixer, or cancel and re-book free of charge within a 24-hour window. That is control without a list to fall back on.

The spine of this project is a flow, not a screen. I designed the end-to-end path that takes a worried homeowner from panic to booked, and mapped the emotional arc before any high-fidelity work so the founders and I could react to a real journey instead of debating in the abstract.

The end-to-end homeowner journey. The emotional curve shows where trust is won and where it dips at the commit moment; the fork at stage 4 tailors the path to three user groups.

The two moments most products skip are the ones I designed hardest for: the match (where trust is either won or lost) and post-purchase job tracking (where booking anxiety needs somewhere to go).

I grounded the design in three inputs:

Competitive teardown

I pulled apart the incumbent marketplaces (the Thumbtack and Taskrabbit pattern) to locate exactly where the journey adds stress, the compare-and-vet middle is where anxious users stall.

Founder and domain insight

Close, fast loops with the founders to pressure-test the single-match bet against how repair jobs actually get booked.

Segmentation data

The signal that users split into distinct groups with different budgets and service expectations, which became the basis for the three-tier model.

User Interviews

I interviewed 6 participants who had recently dealt with a home repair, a mix of renters and owners. They didn't behave like one audience. Three needs-based segments emerged, and they map directly onto the three service tiers I later designed, which is why the journey forks at stage 4 in the User Journey Map.

Tier 1 — Cost-first fixer (budget-conscious renters and owners)

"I just need it fixed without getting ripped off, I don't need anything fancy."

Price anxiety and a real fear of being overcharged
Wants the cheapest reliable option, and will wait to save

Tier 2 — Balanced homeowner (the default)

"I want someone good and fair, I just don't have time to vet ten people."

Decision fatigue from comparing pros
Wants a trustworthy default without overpaying

Tier 3 — Convenience-first owner (time-poor professionals)

"Just send someone great today, I'll pay to have it handled."

Values speed and white-glove service over price
Wants minimal involvement and high reliability

One quote for everyone

What I tried: the founders wanted to keep it simple with a single price for every user.
What happened: the data showed our users weren't one group, they fell into distinct segments with different needs and willingness to pay, so one quote fit no one well.
Learning: tailor the offer to the user, not the average. I brought the segmentation data rather than an opinion, and we shipped a three-tier model, three quotes and three service levels for three user groups.

Leaning too hard into motion early

What I tried: expressive animation using AI for the mascot and splash screen.
What happened: it pulled focus from clarity at the moments that needed to feel calm and trustworthy.
Learning: motion should serve comprehension, not decorate, I pulled it back to where it earns its place.

Screenshot 2026-07-06 at 12.05.19 PM.png

COMPONENTS AND CRAFT

With a tight timeline, I prioritized foundations over a full design system. I established the typographic and iconographic systems, set the brand, and built a component library with variants for each app screen, so the UI stayed consistent and quick to assemble even without a complete tokenized system.

I designed against Apple HIG and Material, and I'm deliberate about where iOS and Android diverge (navigation and back behavior, native controls, touch targets, type). Components included the AI chat and its states, the match card, the tier selector, job tracking, category cards as custom SVG illustrations, and the truck map icons.

MAIN SCREENS

I developed one core solution: an AI-assisted booking flow that matched each homeowner to a single vetted pro and a tiered quote instead of a list to browse, and a human-in-the-loop model that kept users in control, showing the match's reasoning and letting them chat with the fixer, accept a free inspection when the AI couldn't confidently quote, or cancel and re-book free within 24 hours.

As the sole designer, I shipped the core app surfaces end to end:

AI chat - The conversational intake that turns a vague "my sink is leaking" into a structured, quotable job, and sets expectations for an AI product from the first tap.
Home screen - The home base that gives a returning user a clear read on their active jobs and a fast way to start a new one.
Project scheduling - Where the homeowner locks in a time with their matched fixer, turning the decision into a committed booking.
Project and quote elements - The most state-heavy part of the app, 65 states covering every status a job and its quote can move through, so the interface always reflects exactly where things stand.
Messaging - A direct line to the matched fixer, so questions get answered without leaving the app or losing the thread.
User profile - Where the homeowner manages their account details and job history, the layer that makes repeat bookings effortless.

THE MARKETING SITE

Beyond the app, I designed the dual-sided marketing site, with separate entry paths for the two audiences it speaks to, a Fluid Glass-inspired hero, and the stats, USP, and about sections. I made the path from web to app deliberately single: QR-code CTAs rather than scattered download buttons, so acquisition stayed focused. Designing the app and its front door together is what kept the 0-to-1 feeling like one coherent product.

MY LEARNING

Early on I leaned too hard into motion, the mascot and the splash screen, and learned to ease off. At the moments that need trust, motion should help people understand, not just look good. The harder challenge was proving the single-match idea worked with no users yet. I couldn't test my way to confidence, so I built trust into the product itself, with clear states and easy ways out, instead of claiming the bet was proven.

Next up is Voodies II. Previous project NomadNest.