Role
Founding Product Designer
One of two designers. I own onboarding, Explore, profile, premium, and the mentor-side AI flow, plus brand, marketing, and AI production.
Designing the app that turns networking into simply asking for help.
My contribution
I joined Paira as one of its two product designers, before the product had found its shape. I own the surfaces that bring people in and let them help: onboarding, Explore, profile, premium, and the mentor-side AI flow. I also built the brand and every launch material, and I now use AI tooling to tell the product's story.
Timeline
Aug 2025 to present
Concept, beta, pivot, App Store launch, and ongoing growth work.
Team
Early-stage startup
Founders, product, engineering, growth and community, two designers.
Skills
- 0→1 product strategy
- AI interaction flows
- Brand & motion
- Growth & marketing design
- AI-assisted production
I joined before there was a product, and helped decide what Paira should be.
Paira started with one ambition: make it easy for a student or new graduate to reach the one person who can actually help them, without the performance that networking usually demands. When I joined in August 2025 there was an idea and a few screens, so I started close to zero. The direction changed several times on the way: from a swipe deck borrowed from dating apps, to seeing that swiping was not friendly to mentors, to understanding what people actually needed, to request-based matching, an AI plan, and a warm introduction. Each step was designed, reviewed, tested with users, and improved. In a startup this small, one of two designers does many jobs. Mine were to make the idea concrete, to own the surfaces where people enter and give help, to draw the brand, and later to run the launch and marketing creative. A year on, Paira is live on the App Store, 100+ people joined the waitlist and 50+ tested the beta, and the product is still moving, so my work on it is too.
People do not need more connections. They need the right person for one real need, at the moment they have it.
Performative profilesCold outreachUnclear intentNothing to say first
Where it startedA swipe deck borrowed from dating apps. Match first, talk later.
What we learnedSwiping was not friendly to mentors, and a match still left two strangers with nothing to say.
Where it landedStart from the need. Match on a request, plan with AI, open with a warm intro.
The research said the need was real, and that existing tools were built for the wrong moment.
Before the product had a shape, we had to make the case for it. The early research I contributed to named three problems. Cold outreach on the big professional networks feels like shouting into the void. Formal mentors are expensive, scarce, or locked behind selective programs. School mentorship spots are few and run on fixed schedules. The numbers below come from the March 2026 pitch, with their sources.
- 74%
of U.S. Gen Z aged 18 to 25 say they lack access to critical mentorship.
Big Brothers Big Sisters of America & Harris Poll, 2025 - 80%
of professionals believe networking is critical for career success; 70% of jobs are found through networks.
LinkedIn survey, 2017 - 76%
of mentoring-tool users report collaborating with peers and expanding networks online.
Mentorloop, 2025 - 66%
of Gen Z jobseekers build professional relationships through online career platforms.
Handshake survey via Fast Company, 2023
All five wanted to ask for, or offer, career help through something lighter than what exists today.
“I think this is a good idea, but I’m not sure how different it is from [the big networks]. They already allow cold messaging through paid features. The challenge might be making connections feel more reachable and approachable.”
A participant in the first round of interviews
Relational, not transactional. One need at a time, not a whole career.
The big professional and campus networks are broad but transactional: you browse a marketplace of people and send requests into it. Leland and ADPList sell a booked session. Lunchclub and PolyWorks are relational but built around one format each. Paira takes the empty quadrant: relational and multi-purpose, peers helping peers on a specific need, with the introduction handled for you. That is what makes a connection feel reachable, which is exactly what the interviews asked for.
“Find the help. Be the help.”
The line the beta guide launched with in February 2026. Getting career help should feel natural and mutual: ask when you need it, offer when you are ready, without pressure or formality.
We changed the unit of the product from a person to a request, and I redesigned my surfaces around it.
The early Paira borrowed its shape from dating apps: a People tab you swiped through, a Request tab with small bounty posts, and AI promised for later. We shipped that version to beta testers over TestFlight in early 2026. People liked the demo, and it taught us what would make them come back: even after two people matched, they still needed a reason to talk. We moved to a request-based product step by step, through several rounds of design, review, and user testing. The round of user tests below is the one that made the decision clear.
Browsing people stalled.
- Swiping had friction. Each card demanded a decision right away, and that pressure made people quit instead of choose.
- Everyone wanted to know more. In a job search, a headline is not enough to say yes or no to a person.
Real requests got answered.
- A specific need started real conversations. Of the beta’s three features, requests drew the strongest response, on both sides.
- The mentor side needed the same clarity. Helpers swiping through people had nothing concrete to respond to; helpers reading a request did.
A match was not yet a conversation.
- Two people who liked each other still had to invent the first message.
- One-shot AI was not enough. Search understood the intent, then left people on their own. Intent needed a guided path from the ask to the match.
The evidence came from one-on-one think-aloud sessions with six or more potential users, each given the same set of tasks, designed and run by the design team. We brought the findings to the whole team and shaped the new direction together.
What that means in the product.
- The unit is a request
- You describe what you need in plain words. Matching runs on that request, and Paira suggests helpers with the reason each one fits.
- AI does the writing
- It asks two or three clarifying questions and drafts the request. When a helper is chosen, it writes a warm introduction and a short action plan, so the first message is already there.
- Helpers get a short list
- Explore shows a small set of Daily Picks you could answer today. The mentor-side AI flow turns “what can you help with” into matches in a couple of taps.
- Paid means visibility
- Points boost your request for 24 hours. They do not unlock other people.
We made that call together. My part was to carry it into the surfaces I own, and the next section walks through each one.
One loose sentence, two quick questions, a request worth answering.
A replay of the seeker flow the team shipped. Pick an answer to steer it, or let it run.
- Say it the way you would to a friend.No form, no fields. The sentence is the only required input.
- Two clarifying questions, with answers you can tap.Stage and scope are the two things the matching needs; everything else the AI infers.
- The draft appears as editable fields.Title, description, type, industry, location. The AI writes, the person still owns it.
- Post, and see who can help.Helpers arrive with a reason each. The first message is already written.
A request stays legible from posting to conversation.
Paira’s core loop, shared across the team. A post lands in a helper’s Explore, a reply carries a reason, one helper is chosen and the others are told, and Paira AI writes the introduction. Every surface I own sits somewhere on this line.
The pivot made intent visible. It lowered the social cost of asking, and it gave the AI something better than a profile to work with.
I own the surfaces where people enter Paira, find something to help with, and decide to give more.
Five surfaces, in the order a person meets them: onboarding, Explore, profile, premium, and the mentor-side AI flow. Each one had to be rethought when the product moved to requests, and each one shows a different part of that shift.
Collect less up front, so people reach value first.
Onboarding used to be long: many required rounds before a new user saw anything. Early users and retention matter most at this stage, so after research we reordered what we ask for. Now it is a phone number, a code, a name, and a few taps on what you can do and want to do. The rest is filled in later, inside the app, once people have felt the value and want a better match.
Why the AI comes first. The first real screen is “What would you like to do today?” for two reasons. Every new user who states a need adds a real request to the app, which is what helpers come for. And it takes people straight to the core function: say the problem, get a drafted request, see who can help, in the first minute.
What changed. Fewer forced steps, a profile that grows in context, and an AI entry that fills the app with requests while it solves the user’s problem.
A short daily list of requests to help with, so helping feels finishable.
Explore is where a helper finds something to do. Each day it shows a handful of open requests matched to you, with a badge on each that says why it is there. You can save a pick, pull to refresh, and when the day’s picks run out, the screen tells you when the next batch arrives. Tap a pick to read it; “I’m Interested” moves it into your Request Hub.
The question we kept testing was how to show requests to a helper. A feed that never ends makes helping feel like a debt. A short list that ends makes it feel like a favor you can finish today, so helpers come back. And a reason on each card is easier to trust than a score. Three iterations got us there.
Strong MatchQuick WinHigh UpsideTrending NowRecently ViewedBoosted
Iterations 1 and 2First a stream of requests drifting across the screen, then one large AI pick above a grid of small cards. In usability tests the stream packed in a lot but demanded fast reading and felt like pressure to respond; in the grid the small cards were hard to read, and the big card got attention the others did not. We kept the stream for a later release, and decided that until premium launches, every request gets the same chance to be seen.
Iteration 3 · ShippedA short daily list of equal cards, each with a badge that says why it is here, save for later, and a clear “done for today”. We tested it one-on-one with six mentors on the same tasks: a list that ends felt finishable, and a reason on each card was easier to trust than a match score.
What changed. From a swipe deck of bounties to a short list you read, save, and come back to, where every card is something you can act on today, and every request gets an equal chance.
A profile that reads as “what I can help with,” filled in when it starts to matter.
Three constraints shaped the profile. A phone screen is small, so only what helps a decision goes above the fold. A mentee decides in seconds, so the profile leads with what you can help with, what you want to do, and the reason you were suggested. And the decision is about trust, so the numbers a mentee checks sit right under the headline: requests helped, likes received, days on Paira, response time, then recommendations from people you actually helped.
What that gave us. Experience and education are still there, lower down, and the editor breaks them into small steps so nobody stalls on a blank form. “Open to” switches say what you offer, so nobody asks for what you did not. Points, Request Boost, and settings live here too, because this is where a person decides how much of themselves to put in. The profile is never demanded up front; the app asks for it when it starts to matter, which is what lets onboarding stay short without losing match quality.
Make the paid layer about your request getting seen, not about unlocking other people.
Paira’s premium layer runs on points. You earn them by inviting friends who complete a profile, which grows the network, and you spend them to boost a request for 24 hours, to ping someone without a match, or to see who has “Paira’d” you. Boost is the piece I care most about: the one paid action that helps the request rather than the ego, so paying feels like moving your ask forward. The purple sheet and the “Request Boosted!” moment are meant to feel like a small celebration, not a checkout.
Next. Premium Ping and “See who Paira’d you” come from the match era. The next round decides whether they stay or the paid layer narrows to Boost and request visibility.
Two taps from “I want to help” to mentees waiting, so helpers actually show up.
Helpers give the app its supply, and they have less patience than seekers, so the mentor flow had to be the fastest thing in Paira. It opens on suggestion bubbles built from your onboarding intentions and past preferences, sized by how likely each one is to fit, so most helpers reach their matches in two taps. “Tell Us More” is the fuller path: a paragraph in your own words, which the AI reads and turns into tags you can edit before it shows the mentees waiting for you.
The design problem was symmetry without sameness: a mirror of the seeker flow, without its questions. The AI shows its work for a beat and then gets out of the way, and the outcome is a list of concrete things to do, not people to judge. At the end the flow asks one thing, whether to save these answers to your profile, and that is how the mentor flow feeds Explore.
Shown here as core functions, designed with the team rather than owned by me: the seeker-side AI request flow, Request Hub, chat with Paira AI action plans, meeting scheduling, and the helper profile’s “Why you are seeing this person” card.
Five forks, and what each one cost.
Each row records the question we faced, what we chose, what we gave up, and why. None of these was free.
Designed for young, AI-native, and professional at the same time.
The identity is mine: the mark, the icon family, the loading animation, and the type, color, and shape system the app runs on. The component library is shared with the team.
Mark
The wordmark folds into the mark. Tap to replay.
Why it looks this way
- Letters first. The mark grows out of the P and the a in Paira. The yellow frame is drawn as a pictogram, not a letter.
- Two colors, two people. Yellow is the person asking. Violet is the person helping, and the AI.
- Young, AI-native, professional. Acid yellow and violet on near-black feel electric; Playfair in italic keeps it credible; Plus Jakarta Sans does the everyday work.
Type
Find the help. Be the help.
- DisplayPlayfair Display · 56 / 44 / 34 / 26 · 700–800
- HeadlinePlayfair 23 · Plus Jakarta 22 / 18 / 15 · 600–700
- BodyPlus Jakarta Sans · 18 / 15 / 13 · 400
- FootnotePlus Jakarta Sans · 12 · 500 · +2% tracking
- Plus Jakarta Sans · 18 / 16 / 14 · 700–800
Color
Yellow · the person asking#FBFF5C
Violet · the person helping, and the AI#605CFF
Neutral · surfaces and text#131313
StatusOpen · Progress · Chat · Done · Due
Components
OpenIn ProgressCompletedDue in 3 days
Strong MatchQuick WinHigh UpsideBoosted
Post and see the helpersConfirm & See MatchesSkip for now
412162036pill
Motion
- Taps and chips: 150 to 300 ms
- Sheets and screens: 420 ms, ease out, no bounce
- Loading breathes, never spins. AI states show their work for a beat, then leave
The identity had to work beyond the app, so I took it to posters, guides, events, and the website.
From the waitlist in late 2025 to the App Store launch in June 2026, the summer speaker series, and this fall’s back-to-school push, I designed every public-facing piece: recruitment posters, a bilingual beta guide, the beta install site, App Store artwork, launch teasers, and event graphics. In the middle is the newest piece, a concept film made with an AI agent.
With the product live, my work has grown from the screens to the story: bringing in the next wave of students, faster, with AI in the loop.
A launch is a moment; a product is a habit, and habits are built one cohort at a time. This fall’s back-to-school push is the next one. My role has widened to match: the website’s sign-up flows, event and launch creative, and the short films that explain what Paira does in under a minute, now made with an AI agent so each idea reaches students in days, not weeks.
Website & sign-up
Designed the beta sign-up and referral-code flow on paira.org, phone verification to TestFlight in two pages, and the “You are in!” install page that brought in 50+ beta testers before launch.
Launch & events
App Store launch creative, three summer speaker events on breaking into Big Tech, and the referral tier system that ran from the waitlist onward. Live since June 2026, v1.0.1 shipped in September.
AI-made films
Concept films storyboarded and art-directed by me and built by an AI coding agent from our real product assets, like the iPhone Duo film above. Days from idea to cut.
Try the full experience and witness our growth.
Paira is on the App Store and at paira.org. Post a request, or answer one.
A year at Paira taught me two things at once: how to grow inside a team, and how to make the team work better.
Shipping a product from an idea with a small team means the roles blur. I joined as a designer; a year in, I have a much clearer sense of what I can carry, and a few working methods the team now runs on.
How I grew
- From collaborator to shipper. I worked with the PM and the developers on requirements from the first frame to the App Store, and often was the PM, and half a developer, when the work needed it.
- Finding better ways to work together. We moved from everyone working separately and posting updates, to using AI to carry ideas, and to putting the prototype on the web so the whole team could brainstorm and vibe-code on the same screens.
- Defining the product, and my place in it. In mixed roles and an unknown market, I learned to help decide what Paira should become next, and where my strengths are most useful: product judgment, systems, and the craft of the screen.
What I learned at work
- A cleaner path from design to code. As the app grew, three questions kept coming back: where is the latest screen, what is in development, and what is built but still broken. I answered them once instead of every day: a single inventory of current screens, a prototype on the web that anyone can open, and a short rule per role for how a change moves from a meeting to the build.
Team meeting→AI meeting notes, decisions logged→Ticket→Web prototype, one link for everyone→Latest-screen inventory→Per-role sync rules→iOS build
- One design system, app and web. Bringing the app’s system to the web version made the marketing site faster to build, and made this case study possible.
- AI as a working partner. I keep testing where AI helps UI/UX design and where it does not, and I build my own small tools and agents along the way, so my learning compounds with the team’s.