Visa‑Smart Job Match

Inspiration

LinkedIn job‑hunting felt like Groundhog Day:

• 30‑minute scroll marathons,
  
• constant visa‑eligibility guesswork,
  
• manual résumé tweaking for each role.
  

We wanted a one‑click assistant that scrapes only the latest listings, respects OPT/H‑1B constraints, and ranks jobs by true résumé fit— locally and privately.

What it does

1. Initial set‑up

   • git clone → npm install → node server.js
     
   • Upload resume.pdf, visa status, and OpenAI key via a local UI (localhost:3000).

2. Fully automated session

   1. User opens a “Past 24 hours” LinkedIn search tab.
      
   2. Extension/Playwright autoscrolls & scrapes job cards + full JDs.
      
   3. Python service embeds résumé & JDs (text‑embedding‑3‑small), filters visa keywords, and scores cosine similarity.
      
   4. Top N roles appear in the UI and export to top_jobs.csv.
      
   5. (Optional) CSV is e‑mailed via local SMTP.

How we built it

Challenges we ran into

• LinkedIn anti‑bot throttling – solved with randomized delays & session cookies.
  
• Visa keyword chaos – built a YAML taxonomy with positive/negative regexes (“sponsorship available”, “GC holder only”, etc.).
  
• Large PDF résumés – first three pages only, plus embedding cache to avoid repeat cost.
  
• Manifest V3 quirks – needed message‑passing between content script and service‑worker for authenticated fetch.

Accomplishments that we're proud of

• Local‑first privacy – résumé never leaves the user’s machine unless they opt‑in to S3.
  
• 90 % time‑saving – beta users cut job‑screening from ~30 min to < 3 min/day.
  
• Plug‑and‑play – no persistent servers; everything spins up on demand.

What we learned

• Small embeddings are plenty – text‑embedding‑3‑small is 5× cheaper and still nails semantic matching for 30 JDs in <2 s.
  
• UX matters – a single “Run” button and output CSV drove adoption far more than a fancy dashboard.
  
• Scraper resilience – centralizing selectors lets us patch DOM changes in minutes.

What's next for Papa

Chrome Extension → S3 → Lambda pipeline

Built With

• express.js
• javascript
• python