JobSync: AI-Powered Privacy-First Job Search Experience
Overview
JobSync is a conceptual browser extension and dashboard designed to help early job seekers take back control of their search by tracking applications, surfacing follow-ups, and bringing relevant information into the process at the right time, without making the experience feel overwhelming. Research was a team effort. I worked alongside Tanvi, Chinnedum, and Ishrar on the competitive review, 11 in-depth interviews with observation sessions, a 44-participant survey, synthesis, and both rounds of usability testing, and was one of the core decision makers shaping research direction. Shaikh led hi-fi design and prototyping. I contributed to lo-fi framing and ideated design directions for the design kit.
00
Problem
Job searching today means managing dozens of applications across multiple platforms, re-entering the same information into every ATS, and waiting with little to no feedback. Most people start with a spreadsheet, abandon it within weeks, and fall back on starring emails in Gmail just to feel some sense of order. The breakdown happens after you hit apply. There's no visibility, no prompts, no connection between what's landing in your inbox and the applications you're tracking.
Who is affected
Early career job seekers, recent graduates, and career changers applying for their first or next professional role. In our survey, 55% were aged 18 to 24, applying anywhere from 5 to 25 jobs a week, often across multiple job titles and industries. They juggle LinkedIn, Indeed, company websites, Excel, and Notion just to keep up, all of it managed on their own across multiple devices, platforms, and inboxes. Over time, this constant context switching becomes mentally exhausting and leads to burnout. AI-powered job search tools are becoming more common, but many users still approach them with caution. There is interest in automation, but concerns around personal data, privacy, and trust make people hesitant to fully rely on these systems during such an important process. The result is a job search that feels effort intensive but offers very little feedback, making it hard for people to measure progress or feel confident that their efforts are paying off.
role
Primary Researcher, Supporting Designer
team
Kalyani Auti, Shaikh Aziz, Ishrar Islam, Chinnedum Ekeh, Tanvi bhakhar
How might we…
…help early-career job seekers feel more in control of their search through a low-effort, AI-assisted experience that surfaces timely support without adding complexity or compromising trust?
More specifically, how might we
—> Track applications and progress in one place without information overload
—> Reduce uncertainty around outcomes and next steps
—> Provide timely guidance that keeps users proactive
—> Build momentum, visibility, and control during the search
—> Introduce AI with transparency, privacy, and trust
The Design Principles:
Visibility
The single biggest pain point wasn't the volume of applications, it was the near-total lack of feedback after submitting them. Every screen should answer "where do I stand?"
Guidance
Trust
Copying job details, updating spreadsheets, remembering follow-up dates — all manual, all tedious, all prone to falling apart.
Usability testing
We tested the same four tasks twice and watched them improve.
Task
Success R1
—>R2
Error Rate R1
—>R2
Avg. time R1
—>R2
Key change
Onboarding
100% —>100%
42% —>0%
Errors eliminated
3:35 —> 2:39
Step indicator added; condensed to 3 core screens; removed quick tip overload
Save a Job
100% —>100%
42% —> 50%
State change needs stronger affordance
0:57 —> 1:10
Color + animation on save; removed redundant "saved" page
Update status
57% —>100%
100% —> 16%
major improvement
2:03 —> 0:48
Deleted parallel path; added direct status dropdown on tracker cards
AI insights
86% —>83%
43% —> 50%
preselection still creates friction
2:38 —> 1:26
Rewrote prompts to be actionable; added expand/collapse
Reflection
What I'd change, and what I'm taking forward
What would I do differently?
I'd rebuild the AI entry point. In both test rounds, users tried to open the chatbot without selecting a job card first. Color and labeling failed to fix it. The navigation model itself was wrong. I would incorporate the insights inside each job card, not on a separate page.
Where did the research fall short?
Our 44 survey respondents skewed young. 55% were aged 18 to 24. That limited what we could claim about mid-career seekers like Rafael. We also tested a simulated prototype instead of a live extension. Friction around the save animation and email integration stayed invisible as a result.
What would I test next?
I would test an interview prep mode that triggers when a status changes to "Interviewing." Company research, shared interview experiences, role-specific prompts. One open question would drive the test: do people use these at the moment they need them, or does more content overwhelm them at the most stressful point in the search?
What did I learn?
Staying open when research contradicts you takes real effort. We assumed more features and more AI would reduce stress. Users asked for fewer features that worked reliably, with their privacy protected. We rebuilt around what they said.
I understood "build what users need, not what you can build" long before this project. But the clearest lesson came from watching a user struggle with a feature we were proud of. It confirmed what the research already told us.