JobHunt: What I Was Actually Trying to Fix

My previous project, SalesCall, came from watching an employee try to handle cold calls on a device that couldn't run the tools available. The apps that existed were just too heavy. The product decision was simple: build something lightweight and focused entirely on that one workflow.

JobHunt followed the same logic, but the problem was bigger. I kept noticing how fractured the modern job search had become, and I couldn't stop thinking about why nobody had fixed it.

The Problem

Most people job hunting are switching between LinkedIn, Wellfound, various Slack communities, and whatever job boards are relevant to their field. Then they're managing the actual work in a mix of spreadsheets, browser tabs, and half-finished Google Docs.

On top of that, every application asks for roughly the same effort. Rewriting a CV to match the role. Writing a cover letter that doesn't read like a template. Tracking where each application is in the pipeline. Preparing for interviews without a clear framework to work from.

The frustration wasn't that job searching is hard. It's that the tools force you to do the same repetitive work over and over, across five different places, with no memory of what came before.

I decided to treat it as a product problem. Could I consolidate those scattered pieces into one place and make the AI do the repetitive work, so the person applying could focus on the parts that actually require them?

The Build

I built JobHunt in my spare time using Next.js, Supabase, and Gemini AI, keeping the stack lean and running almost entirely on free-tier infrastructure. The goal wasn't to build something impressive for its own sake. It was to prove that a genuinely useful tool could be built this way.

What's in the platform right now:

• A unified job feed pulling from eight boards, including Remotive, RemoteOK, and Adzuna, plus a paste-a-job flow for anything not on an integrated board

• AI CV tailoring with a split-pane diff view so you see exactly what changed before you save anything

• Cover letter generation that actually sounds like a person wrote it, with style and length options

• Interview prep pulled from the actual job description, with full STAR model answers written from your own CV

• A full application pipeline tracker from draft to offer

• Gmail OAuth to send the application without leaving the app

Where the Real Work Was

The most interesting parts of building this weren't engineering problems. They were product decisions.

One early one: what does a user see the moment they log in before their feed has populated? A blank state is a trust problem. Getting that right, making the zero state feel intentional rather than broken, took more thought than most of the features.

Another one: CV text extraction. If the AI is going to tailor your CV or generate STAR answers from your background, it needs clean input. Poorly encoded PDFs produce garbage output. I had to build around that before any of the AI features would actually work reliably for real users.

These are the details that don't show up in a feature list but are where the actual experience is won or lost.

What's Next

JobHunt is in beta and open for testing. I'm looking for feedback from people who are actively job hunting, from product and engineering folk who want to dig into the workflows, and from anyone who has opinions on where this kind of tool should go next.

If any of that is you, try it and use the built-in feedback tab. I read everything.

JobHunt is currently in beta. Try it and tell me what you think.

Request Early Access