xRADAR
A Mac app that finds the jobs that fit you and writes a CV for each one.
Why I built it
Search and matching
Jobs are scattered across many sites, so looking for a product role meant checking each one. The same job appeared in several places. I'd forget where I'd applied and read the same openings again. Most didn't fit, yet they stayed in the results.
CV and cover letter
Every application needs a tailored CV and cover letter, and that takes time. I wanted one base of my experience and a quick way to build both for any job, using only facts from that base.
How it works
- Set the roles you want and the location, and add your CV and past work to your experience base.
- The app collects new jobs from job sites and company career pages into one list, without duplicates or jobs you've already seen.
- It selects the jobs that fit you.
- For the job you choose, it builds a CV and cover letter from your experience base.
- You track your applications in the same list.
Screenshots
Features
Search
- Jobs from LinkedIn, Just Join IT, No Fluff Jobs and Solid.jobs.
- Company career pages on Greenhouse, Lever, Ashby and Workable.
- Removes duplicates across sites, with AI checking the unclear cases.
- Remembers every job, so you never review one twice.
Matching
- Sorts jobs by your request into three groups: strong fit, possible fit and no fit.
- A match score from 0 to 100 with an explanation.
- How your role and industry fit, what's missing, and advice on what to watch for.
- A short summary of every job.
Company profile
- What the company does and how it is doing, including its financial position.
- Employee reviews, with sources and dates.
- Red flags, if there are any.
- A summary of how the company fits your goals.
CV and cover letter
- Works through your Claude or ChatGPT subscription, with no extra payment.
- Built from your experience base, in a chat right inside the app.
- An AI critic checks each CV as a hiring manager would, and against your facts.
- If something is missing, the app asks you and saves the answer to your base.
- A clean PDF, with the layout checked by code.
Tracking
- Mark jobs as Interesting, Not for me or Applied. Jobs you hide stay hidden.
- Follow each application through Interview, Offer or Rejected.
Key decisions
- AI answers, code decides. Without clear rules an AI model gets noisy: the same job can get a different answer each time, so a good one could vanish by chance. So it answers simple questions about the role, level and location, and fixed rules in code decide the fit. Results stay stable, mistakes are easy to find, and the cheapest model is enough.
- For a CV, choosing beats writing. Bullets written by a model came out long and generic, and sometimes it started making things up. So I built a set of ready bullets from my real experience, with versions for different angles, because the same work can be presented differently for different jobs. It took work up front, but now the model only picks the right ones and code assembles the CV. A draft went from about 3.5 minutes to 30 seconds.
- Test on one, then run on many. One bad prompt on thousands of jobs wastes hours. So a bulk run can't start until a test on one job has passed. That rule is in the code, not a promise.
- From my tool to anyone's. The first version had my own rules in the code: my roles, my city, my name. As soon as the idea proved useful, I moved all of that into settings and gave the app a blank starter profile, so someone else can use it for any role. Packaging and sign-up come later.
What's next
Next I want to find out if this helps other job seekers, not only me. First comes a simple setup and a Mac download, so people don't need developer tools. The test: do their applications turn into interviews? The app already tracks this.
Tech
- Collection
- Python, python-jobspy
- Data
- SQLite
- Models
- Claude Code CLI or Codex, through a subscription
- Interface
- HTML and plain JS
- Mac app
- Swift, WKWebView
- CV
- Typst templates to PDF