ayushnau/workday_jobautomator

This project is a Chrome browser extension that automatically fills in job applications on Workday, the hiring platform used by many large companies. The user uploads their resume once, and when they open a job posting on a Workday site, the extension reads their resume and types the answers into the form for them. The AI part can run locally for free using Ollama, a tool for running language models on your own computer, or it can call Google's Gemini API instead. Setup is done with a few shell scripts. One script downloads two helper libraries called PDF.js and mammoth.js, which are used to pull plain text out of PDF and DOCX resumes. A Python script generates the icons, and a build script produces a folder called dist and a zip file. To install in Chrome, the user opens the extensions page, turns on Developer mode, and picks Load unpacked, because Chrome blocks drag-and-drop installs from outside its Web Store. To run with Ollama, the README shows the commands to start the server with cross-origin requests allowed and to pull the llama3.2 model. The flow runs in phases. First, the popup reads the resume file as raw bytes, extracts text, and sends it to a background service worker. The service worker asks the language model to turn the text into a structured JSON record with name, contact details, work history, education, and skills, and saves it in Chrome's local storage. Second, when the user clicks Start Autofill on a Workday page, a content script on the page loads the saved resume data, shows a floating panel on the right, and starts walking through the form step by step. For each field the extension first tries fast pattern rules for common questions like name, email, phone, GPA, salary expectations, or notice period. When a field is unknown, it sends the field label and the resume to the AI and uses the model's answer. Different handlers cover text inputs, native dropdowns, Workday's custom dropdowns, radio buttons, checkboxes, date masks, and typeahead search fields. The README is honest about what cannot be done: file uploads and CAPTCHAs need manual handling, and some company-specific questions may still get a wrong answer that the user has to review. The layout of the source is documented in detail, with separate folders for the popup UI, the background service worker, the resume parser, AI client code for Ollama and Gemini, and a content folder split into sections for personal info, experience, education, languages, skills, screening questions, and open-ended questions. A demo video link is included at the end.