What if recruiters could see who a candidate really is before the first call? Peeks explored a video-first discovery experience that treated people as more than a list of keywords on a resume.
Cvent event schedule app. Lead product and motion designer.
Hackathon project, wedding planning app. Lead product designer.
Nike enterprise events calendar. Lead product designer.
Context
For decades, hiring has started with the same artifact: a document. Resumes, cover letters, LinkedIn profiles. All text, all static, all optimized for keyword matching rather than genuine human understanding. Recruiters scan hundreds of these documents per week, trying to extract signal from formatting and bullet points.
The problem is obvious to anyone who has hired before. The best candidates are often terrible on paper. The people who light up a room, who communicate complex ideas clearly, who bring energy and curiosity to a team, none of that shows up in a PDF. Recruiters know this. They talk about "culture fit" and "soft skills" constantly, but their tools give them zero visibility into those qualities until the first interview.
Meanwhile, an entire generation of talent communicates through short-form video every day. They are comfortable on camera, expressive, authentic. There was a gap between how people naturally present themselves and how the hiring process asks them to.
Discovery & Research
This project started with a simple question: what do recruiters actually look for when they evaluate someone? Not what they say they look for in a job description. What actually moves the needle when they decide to advance a candidate?
R12Keywords get a resume into the pile. How someone explains a tradeoff gets them the job.
QWould you watch a short video?
R12If it’s under a minute. I don’t have time for more than that.
QAnything that worries you?
R12Legal will ask about bias before you finish the sentence.
QWhat tells you to move someone forward?
R07Honestly, I know in the first thirty seconds. How they frame the first answer. The rest of the call is me confirming it.
QWhat are you listening for?
R07Clarity. Can they get to the point. Then energy, whether they care about the work. None of that is on the resume.
QA short video instead of the call?
R07Faster, sure. But you’d see age, you’d see race. That’s a bias problem before it’s a product.
under 35 ✓✓✓✓✓✓✓ ✗
35 and over ✓✓✓ ✗✗✗
“only if I can retry”
Affinity mapping from 100+ interviews, clustering evaluation criteria and behavioral patterns
Strategy & Framing
The research pointed to a clear opportunity. Recruiters were spending enormous effort trying to infer personality, communication skills, and culture fit from text-based profiles. What if they could just see it? Short video "peeks" that let candidates show who they are in 30 to 60 seconds, and let recruiters browse those clips the way they browse a social feed.
We framed the concept around three principles that came directly from the research.
Recruiters evaluate fast. Any video experience needed to match or beat the speed of scanning a text profile. That meant short clips, instant playback, and a swipe-to-advance interaction that felt effortless.
We did not want polished video resumes. We wanted authentic moments. The prompts and format had to encourage natural responses, not rehearsed pitches. Think "tell me about a problem you solved recently" rather than "describe your five-year career plan."
Video inherently carries more demographic signal than text. We built bias awareness into the design itself: structured prompts, blind-first options where recruiters see the text profile before unlocking video, and clear guidelines for evaluation criteria.
Design Process
The core design challenge was translating the speed and addictiveness of a social media feed into a professional recruiting context. It needed to feel familiar and fast, but also intentional and respectful of the people on screen.
We started with the feed. Vertical video cards, full-bleed, one candidate at a time. Swipe up to advance, swipe right to save, swipe left to pass. The interaction language borrowed from dating apps and short-form video because users already understood it. No tutorial needed.
Each card layered video with a minimal text overlay: name, current role, years of experience, and a one-line AI-generated summary. Tapping paused the video and expanded the full profile underneath. The goal was to let recruiters stay in the flow state of browsing while giving them a quick path to depth when they wanted it.
The other half of the design was the candidate side. We needed people to record short clips that actually revealed something useful, without making the process feel like an audition. We designed a prompt-based recording flow: candidates pick from a set of questions, record a 30 to 60 second response, preview it, and submit. No editing tools, no filters. Just a camera and a question.
Early testing showed that specific, low-stakes prompts produced the best clips. "Walk me through your morning routine at work" outperformed "Tell me about yourself" by a wide margin in terms of authenticity and recruiter engagement.
felt authentic
A ✓✓ · B ✓✓✓✓✓✓✓
watched to the end
A 38% · B 91%
We built a functional prototype in Figma and a lightweight native prototype for mobile testing. Over three weeks, we ran the prototype through five rounds of testing with recruiters and three rounds with candidates. Each round was small, five to eight participants, but tightly focused on specific questions about speed, comfort, and signal quality.
Key Design Decisions
Early in the project, we debated whether video should replace the text profile entirely. Testing made the answer clear: no. Recruiters wanted both. They wanted to scan a text profile for hard qualifications, then watch a clip to gauge the human qualities. So we designed Peeks as a layer on top of existing profiles, not a substitute. Text first for filtering, video for evaluating.
Jamie C.
Alex K.
Maya L.
Taylor R.
Omar N.
Traditional sourcing is active. You type a query, scan results, refine, repeat. Peeks introduced a passive mode: an AI-curated feed of candidates matched to your open roles, served one at a time. Recruiters told us this felt less like work. They compared it to discovering music on a streaming playlist versus searching a catalog. The feed did the heavy lifting, and they just reacted.
We could not ignore the bias question. Video reveals age, race, gender, and physical appearance in ways that text does not. Our approach was multi-layered. Structured prompts ensured every candidate answered the same types of questions. A "blind first" toggle let recruiters review text qualifications before seeing any video. And we designed evaluation rubrics directly into the interface, nudging recruiters to score on specific criteria rather than gut feeling alone.
Outcomes
Peeks was a concept and research project, not a shipped product. It never went to production. But it was never meant to. The goal was to explore a fundamentally different approach to candidate discovery and pressure-test it with real recruiters. On that front, it delivered.
The strongest signal was recruiter enthusiasm. People who had been skeptical about video in hiring became advocates after using the prototype. They said things like "this is what a phone screen should be" and "I can tell more from 30 seconds of this than 30 minutes of reading resumes." The concept validated that video has a real place in recruiting, but only if it is fast, structured, and designed with bias in mind.
Several insights from Peeks later influenced the core Wellfound product roadmap, particularly around richer candidate profiles and AI-powered evaluation. The project proved that innovation work within a product org does not need to ship code to create value.
Reflection
Running a skunkworks project inside a company that is also shipping a core product is a balancing act. You need enough independence to think differently, but enough connection to the main team that your insights can flow back. Peeks worked because we treated it as research with a prototype, not a stealth product launch. The framing mattered.
The biggest lesson was about the role of prototypes in building conviction. Slide decks and research reports are easy to nod along with and then forget. A working prototype that puts a recruiter in front of actual video clips and lets them swipe through candidates creates a visceral reaction that no document can match. Several stakeholders who were lukewarm on the concept became champions after a five-minute demo.
I also learned to design for the risks, not just the opportunities. The bias question was the hardest part of this project, and addressing it head-on made the concept stronger. Ignoring it would have made it dead on arrival. The best innovation work does not just imagine what is possible. It honestly confronts what could go wrong.