Step one
Describe the role you are tracking
Paste a job order, a candidate resume, or a plain description of the work. The text is embedded once and the vector stays in this browser tab. Everything after that runs locally on your machine.
First scan downloads the job index, about 60 MB. It is cached after that.
Step two
Nearest clusters
Each card is a cluster of near identical postings, closest match first. Keep pulls that cluster down to this tab. Skip drops it and teaches the model what to rule out.
Step three
Ranked openings
J keep K drop Space next model refits on every keystroke