Our work / Hello Job Tracker
Hello Job Tracker: AI help built not to invent your experience
A free job application tracker that reads job ads, estimates how well each one fits you, and drafts tailored resumes and cover letters.
The problem
Tailoring every application takes hours, and general AI tools will happily add experience you don’t have. That is easy to miss, and it’s the applicant who pays for it in an interview.
What we used to build it
- Interface
- React, Vite and Tailwind CSS, in TypeScript
- Server
- Hono on Node.js, hosted on Fly.io in Sydney
- Data
- Postgres with pgvector on Neon, through Drizzle
- AI
- Anthropic’s Claude models through the Vercel AI SDK, with spending limits on every call
- Reading job ads
- A page reader, plus a bookmarklet that works in your own browser
- Edge and storage
- Cloudflare for traffic, attack protection and file storage (R2)
The choices that make it responsible
Every line has a source
Each line in a draft names the resume line or fact it came from, and software checks that source exists. A draft that fails goes back once. If it fails again, it isn’t saved.
Fit scores show their reasons
Each score breaks into six areas, with the line from the ad behind it. A deal-breaker you set caps the score, however good the rest looks.
You stay in charge
It never applies for a job and never sends an email. You decide what goes to an employer.
Costs and keys are controlled
Every model call counts against a limit. A visitor’s own API key is held in memory only and is never stored.
What we’re still improving
Evaluations now cover fit scores, grounding, job reading, and fairness in fit scores and drafts. Interview preparation, written answers, follow-up messages and the chat are still to come.
Want something like this for your organisation?
Tell us what you’re trying to do. If AI isn’t the answer, we’ll say so.