https://github.com/cloudhighfive/alphajob
I began working on a new project which targets the developer problem of handling endless job application tasks. The process of completing forms on Lever, SmartRecruiters, and Greenhouse requires full-time dedication because it takes up that amount of time. I began developing this solution during the previous day.
The solution requires advanced techniques for web scraping and browser automation to solve this complex technical problem.
Code Strategy & What's Done So Far:
The system uses Playwright for browser control to enable smooth navigation between different ATS platforms through a Python-based framework.
The system uses separate modules for each major platform including Ashby and Workable to handle URL and form field extraction tasks.
The system successfully detects and fills all typical application fields which include name, email, and basic work history information.
The Current Grind (Technical Issues Requiring Refinement):
The system faces two main challenges because it triggers aggressive spam filters which Playwright's built-in stealth features fail to prevent. The system needs to learn how to hide headers and create fake human-like submission delays to defeat complex bot detection systems.
The selection process for gender or demographic options proves to be a major stability issue because it seems like a basic operation. The different application forms use inconsistent HTML tags and hidden containers and unreliable CSS selectors which makes my element identification system unstable.
Top comments (1)
On the tech side, lean hard on role/label/ARIA locator, handle virtualized lists/iframes/shadow DOM with real readiness checks, and ship a pluggable ATS adapter backed by contract tests and fixture snapshots. For stability/UX, scope radio and checkbox groups by their legend, confirm state after clicks, use an “uncertainty score” to pause only when things get weird. Security and ops-wise, keep PII local-only and encrypted, log via an allowlist with redaction/hashes, offer GDPR/CCPA export/delete, surface per-ATS success/failure and “needs-human” metrics, use token-bucket with jitter/cool-downs, run headful for assisted mode and headless with tracing on schedules, prioritize Workday/Taleo/iCIMS, and give folks a desktop app or extension.