Screen remote jobs against a timezone and skills profile
At a glance
A job seeker uses JobKit to extract posting requirements with local Gemma and inspect the evidence behind the screening result. This workflow is listed under job-search, remote-work, evidence-checking.
- Agents / tools
- JobKit, Gemma 3 4B, Ollama
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) Install stable Rust with Cargo and local Ollama, then run `ollama pull gemma3:4b`. The author developed the tool on Ubuntu in WSL. 2) Clone https://github.com/Lymah123/jobkit and enter its directory. Edit profile.toml with years_experience, tz_offset, and your skills list. 3) Save a job posting as a local text file. Run `cargo run --release -- analyze posting.txt --rules-only` for location, experience, and timezone checks without the model. 4) Run `cargo run --release -- analyze posting.txt` to add Gemma skill extraction. Rust verifies proposed skill terms against the posting and supplies evidence lines. 5) Read those lines and any Review flags before deciding whether to apply. Treat the output's distinction between required, optional, and merely mentioned skills as part of the review.
How it went
The author measured model-assisted processing at 43–143 seconds per posting on their machine.
Worth knowing
The author did not measure apply-or-skip verdict accuracy. Long skill lists can be under-extracted, and location boilerplate can trigger review flags.
