Project In Progress
JobSort.ai
█
Development started: June 2026 | It is completely unfinished, but it is being built right.
What this is.
JobSort is a locally hosted, open-source tool that does one thing: it finds job listings, strips out the junk, and sorts them so you only see what actually fits. That's it. No auto-apply bots. No networking dashboards. No telling you to "optimize your personal brand." Just jobs worth your time.
Right now, I am focusing on perfecting this pipeline for just one small, tight set of job posting sites. No massive, broken firehose. Just making it work flawlessly on a few sources first.
Why It's Not Another Garbage Wrapper
Most "AI" apps are literally just a basic text prompt sent to OpenAI. They are lazy, expensive, and they hallucinate. it filters out the garbage using an ML model running directly on your machine (no cloud, no API cost, no one scraping your resume), and by the time a job hits your screen it has already been scored against your actual background. Not keyword matching. Math.
An LLM only gets involved after the math already decided the job is worth looking at. That is how you keep it honest and keep it cheap.
Your Data. Your Models.
Because this runs locally on your machine (built natively with Rust and Tauri), your data stays yours. Your resume, your search history, and your application notes stay on your own hard drive. No cloud databases are scraping you, and nobody is selling your info.
You plug in whatever language model you want to use and swap it out whenever you feel like it. You control the hardware.
What pisses you off the most?
I built this because I hit every one of these problems myself. Before I accidentally solve the wrong thing, I want to know what is actually broken for other people.
The AI that rewrites your resume until it sounds like a LinkedIn post from a bot. The tool that charges you $40 a month to organize your own rejection letters. The one that calls itself AI and is literally just a ChatGPT prompt with a logo slapped on it.
Tell me. I will read every single one.