PERSONAL TOOLS · 2026
The Pipeline
- My contribution
- Automation design & engineering
- Current chapter
- Runs daily · two stages of review
- For
- Personal project
Less repeated reading.
More promising possibilities.
Useful job-search work was getting buried under unsuitable listings and missing details.
The Pipeline makes a deliberate split: a cheap first pass prepares the candidates; a separate Claude workflow does the deeper reading.
A clearer shortlist, with less repeated reading.
THE NEED
The search was becoming its own job.
Searching for work meant reading too many unsuitable roles, then keeping track of the ones that might fit. The automation needed to save time without creating a costly second job.
I built the Python collection and screening workflow, connected email alerts and Trello, and refined the filters against problem listings. The deeper research stays in the existing Claude workflow rather than being duplicated through paid API calls.
02 · THE MAKING
Know what to check, and when to stop.
Gather
Listings arrive from job sources and email alerts, including sources that are awkward to crawl directly.
Check
The first pass checks titles, company identity, location, remote eligibility, and pay. Missing details stay visibly uncertain.
Organize
Relevant candidates reach Trello, with duplicate checks and readable details to support the next review.
Review
A separate Claude project handles the deeper reading of job descriptions, employers, and the actual fit.
WHAT WORKS TODAY
A clearer shortlist for the deeper reading.
- Daily collection and first-pass screening prepare candidates for a separate Claude review.
- In one validation replay, screening reduced 25 candidate cards to 13 while preserving both confirmed keepers. This is a bounded test result, not a claim about every run.
- Unclear details stay visible. Private runner records and personal application history are excluded from this portfolio.
BUILT WITH
Python · Gmail · Trello · GitHub Actions · Claude
Last reviewed September 2026