ProjectsJob Hunt Automation

Local Python automation

Job Hunt Automation

A daily local assistant that gathers approved job feeds, applies deterministic fit and deduplication rules, and prepares a Google Sheets queue for manual review.

Role
Sole designer and developer
Period
2026
Stack
Python · Google Sheets · Windows Task Scheduler · Himalayas · Jobicy

Evidence gallery

01 / 03

01 / Controlled local capture

Dashboard

October 7, 2026 dashboard capture from a local run; counts are a point-in-time queue, not hiring outcomes.

Optional explainer

See the workflow at a glance

This animation explains the workflow; it is not a recording of a live run. The dashboard captures are real, and applications remain manual.

Job Hunt animated overview poster with the message “Too many listings.”

Animated overview

This animation explains the workflow; it is not a recording of a live run. The dashboard captures are real, and applications remain manual.

02 / Challenge

Finding relevant openings should not consume the time needed to evaluate them.

Repeatedly checking job boards creates duplicates and context switching. I wanted a local assistant that prepares a consistent review queue without deciding where to apply or contacting employers on my behalf.

03 / Approach

A scheduled Python run applies explicit rules before writing to Sheets.

A daily Windows run collects listings from the approved Himalayas and Jobicy feeds, normalizes them, applies deterministic fit rules and deduplication, and updates a Google Sheets workbook. The dashboard, technical queue, and pipeline make the next human action visible.

  • No automatic application, message, or outreach is sent.
  • Optional Gemini and Ollama paths are disabled, not active production AI.
  • The workbook supports manual review, status tracking, and follow-up.

04 / Result

The local workflow makes the search reviewable while keeping the final decision manual.

The October 7 screenshots show a working local dashboard and queue. They are not evidence of interviews or offers. The architecture diagram explains how data reaches the workbook and where automation stops.

What I would improve next

  • Continue reviewing feed quality and rule drift as sources change.
  • Keep manual decisions explicit if new integrations are added.