Cut a batch of job postings down to the real roles that fit your resume and flag the ones that offer work-permit sponsorship

# Cut a batch of job postings down to the real roles that fit your resume and flag the ones that offer work-permit sponsorship All postings and employers are synthetic (employers are neutral codes). Inputs: `resume.json` (role_families, seniority_min, seniority_max, locations), `period.json` (as_of date) and `postings.csv` (posting_id, employer, title, role_family, seniority, location, posted_on, description). Seniority runs junior < mid < senior < staff. `recruiter-notes.txt` is a recruiter's checklist; where it disagrees with the rules below, the rules win. Rules: - Duplicates: postings with the same employer, the same title after lower-casing and collapsing runs of spaces, and the same location are one role. Keep the copy with the earliest `posted_on` (ties: the lowest posting_id); every other copy is a duplicate. Same title at a different location is a different role. - Stale: a kept posting is stale when `as_of` minus `posted_on` is more than 30 days. Exactly 30 days is not stale. - Fit: a kept, non-stale posting fits when its role_family is in the resume's role_families, its seniority is between seniority_min and seniority_max inclusive, and its location is one of the resume's locations. These are the real roles. - Sponsorship: a real role offers work-permit sponsorship only when its description has a line that is exactly `Sponsorship: available` (letter case aside). `Sponsorship: not available` and `Sponsorship: none` mean no. - Write `out/shortlist.json`: `{"real_roles": [<sorted posting_ids>], "duplicates_dropped": [<sorted posting_ids>], "stale_dropped": [<sorted posting_ids of kept postings that are stale>], "sponsor_roles": [<sorted posting_ids>], "roles_by_employer": {"<employer>": <number of real roles>, ...}}`. - Roll-ups: Employers in postings.csv with no real roles are left out. ## How this run works - Work on your own: nobody will answer questions in this task, and writing to `ask/` counts against the run. - Write only under `out/`. Create `out/` if it is missing. Inputs are synthetic; don't fetch anything from the network. - Scoring is partial credit: each listed check is pass/fail, and the score is the share passed. Numbers are compared to within 0.005 unless the task says otherwise.