Crowdsourced Resume Library

Writing a resume that stands out is mostly guesswork. Students get plenty of advice and templates, but almost no proof of what actually gets past a recruiter’s screen.

So I teamed up with five software engineer friends to build Resumes.fyi: a library of crowdsourced resumes that passed the screening process at more than 140 tech companies. As co-founder and head of product and design I owned product strategy, and as the only designer I owned everything from branding and identity to interaction design and usability testing.

A crowdsourced library has a built-in catch: nobody browses an empty shelf, and nobody shares to a shelf no one visits. Every design decision that follows comes back to that loop: bring people in, make it easy to give back, and make what they find worth returning for.

Before designing anything, I wanted to confirm this was a real problem. We surveyed 584 people about job applications and resume prep, then ran 25 in-depth interviews with university students, recent graduates, and experienced professionals, over Microsoft Teams and in person.

The answer kept coming back the same way: people didn’t want more advice. They wanted real examples, from people like them, at the companies they were aiming for. We turned that into three personas, and each one shaped a part of the product.

Finding the gap

Platforms like Cofolios and Levels.fyi share a similar spirit, sharing real portfolios and real salaries. None of them focused on the resume screen itself. That was our opening.

We marketed to people looking for a job, mostly university students. Studying other people’s resumes to figure out what’s wrong with your own can feel tedious, even a little discouraging, and I didn’t want Resumes.fyi to feel that way. So I gave it color: the language of studying, with sticky notes, paper textures, pins, grids and stickers, set in a palette of seven bright colors that keeps the experience light and a little playful.

The identity helped us land 1,000 registrations during the waitlist period before beta.

Every new visitor was a potential contributor, so the upload flow had to feel effortless. Early internal feedback said the first version was too long: it felt like filling out a job application. People also worried about legibility, since there was no way to zoom in on a resume.

I cut the flow down to three steps, made anything non-essential skippable, and added a live, zoomable preview so people could see exactly what they were sharing as they went.

More resumes only matter if people can find the right ones. The explore page, where every uploaded resume lives, had to show what matters without opening a single file: experience level, role, and the companies where the resume passed the screen.

I rebuilt the page from its parts, starting with the pieces people touch most.

Putting the pieces together

Here’s where those pieces come together. The sidebar anchors the left, a result count and sorting sit up top, and the cards fill a grid you can scan in seconds. We A/B tested this layout against the original page with 40 participants, and 82% preferred it.

Once people could find resumes, the next question was obvious: what does a given company actually look for? Company pages answer that directly. A directory lists every company in the library, and each company page summarizes the resumes that passed its screen, from common certifications and open-source work to average GPA, past jobs and project count.

As the library grew, Resumes.fyi became an AI recruiting platform. I designed insight dashboards that synthesized more than 600 user interviews into role and candidate fit benchmarks, and developed the ranking logic behind AI resume-to-role matching.

In December 2025, Resumes.fyi was acquired by Aced. My designs carried over into Aced’s product, putting Resumes.fyi in front of Aced’s full user base: more than 100,000 users, including recruiters at Vercel and Pinterest.

What I learned

1. 01
   Solve the supply side firstA library is only as good as what people share. Shortening the upload flow lifted submissions 52%, and everything else depended on that.
2. 02
   Let evidence settle itAn A/B test, not opinions, decided the explore page.

What I’d do differently

I’d put real students in front of the upload flow before launch, not after. Internal feedback caught the friction, but testing earlier would have saved us a full redesign.