🔍 Job Scout — AI-Powered Job Screener for Product Managers

Local job screening system for Product Manager positions. Aggregates jobs from 12+ sources simultaneously, applies a hybrid scoring system (deterministic keyword matching + Groq LLaMA quality bonus), and ranks results so you can focus on what matters.

Built by a PM, for PMs — specifically tuned for senior remote roles in AI/LLM, fintech, and enterprise B2B.

---

Quick Setup

First time

# Backend
cd backend
npm install
npx playwright install chromium
cp .env.example .env
# Edit .env with your API keys (at minimum: GROQ_API_KEY and APIFY_TOKEN)
node server.js

# Frontend (new terminal)
cd frontend
npm install
npm run dev

Already installed

git pull

# Backend
cd backend
node server.js

# Frontend (new terminal)
cd frontend
npm run dev

Opens at http://localhost:3000

---

Job Sources

Source	Method	Key Required
LinkedIn	Apify Actor (curious_coder/linkedin-jobs-scraper)	✅ APIFY_TOKEN
Remotive	Public API	❌
Himalayas	Public API	❌
WeWorkRemotely	RSS scraper	❌
Wellfound	Playwright scraper	❌
Jobspresso	Custom scraper	❌
WorkingNomads	Custom scraper	❌
EuroRemoteJobs	Custom scraper	❌
JSearch (Indeed + Glassdoor + more)	RapidAPI	✅ JSEARCH_API_KEY (200 req/month free)
SerpAPI (Google Jobs)	API	✅ SERP_API_KEY (100 req/month free)

LinkedIn jobs without a full description are automatically enriched via JSearch job-details — only for relevant titles, to preserve the monthly quota.

URL validation runs on slug-based sources (Himalayas, Wellfound, etc.) to filter expired listings before scoring.

---

Getting API Keys

Key	Where	Cost
GROQ_API_KEY	https://console.groq.com	Free
APIFY_TOKEN	https://apify.com → Settings → Integrations → API tokens	Free tier: $5/month credit (~5,000 jobs)
JSEARCH_API_KEY	https://rapidapi.com → search "JSearch by OpenWeb Ninja"	Free: 200 req/month
SERP_API_KEY	https://serpapi.com	Free: 100 req/month

---

How Scoring Works

Hybrid scoring — two independent layers:

Layer 1 — Deterministic base score (code only, no LLM)

Scans title, description, and location for verifiable signals:

• Domain keywords: LLM, MCP, AI Agents, Open Finance, Fintech, Payments, Enterprise B2B, Observability, Azure
• Title match: Head of Product, Senior PM, Technical PM, Principal PM, etc.
• Location: EU-based (+15), Worldwide (+8), UK/CET (+8)
• Salary disclosed (+5), seniority explicit in title (+5)
• Relocation/residency requirement penalty (−15)
• Cap: 75 points

Layer 2 — LLM quality bonus (Groq LLaMA 3.1, 0–25 points)

Evaluates qualitative fit that code cannot measure:

• Actual product/problem domain fit
• Company stage and growth potential
• Role complexity and leadership scope
• Red flags not caught by keywords

Final score = baseScore + qualityBonus (max 100)

Recommendation thresholds:

• ≥ 75 → STRONG FIT
• ≥ 55 → GOOD FIT
• ≥ 35 → CONSIDER
• < 35 → SKIP

---

Automatic Filters

Applied before scoring — no tokens spent:

• Junior / Associate / Intern titles → SKIP
• On-site / office-only roles → SKIP
• US-only roles → removed (checks patterns, US states, US cities in location)
• Expired links → filtered via HEAD request validation (slug-based sources only)

---

How to Use

1. Run Scan — fetches from all sources, filters, scores, ranks
2. View rankings sorted by score — each card shows baseScore + qualityBonus breakdown
3. Expand a job to see highlights, matched skills, red flags, and LLM summary
4. Click Mark as Applied on jobs you apply to
5. Track status in the Applications tab (applied → interview → offer / rejected)
6. History tab loads any previous scan with full details

---

Customize Your Profile

All search criteria live in backend/config.js:

• PROFILE — accepted/blocked titles, skills, deal breakers, salary targets
• CANDIDATE_PROFILE — your background, used by the LLM scorer to evaluate fit
• LINKEDIN_SEARCH_QUERIES — keywords and locations for LinkedIn via Apify (25 queries by default)
• JSEARCH_QUERIES — queries for JSearch/RapidAPI

Keyword weights and scoring rules are in backend/scorer.js → DOMAIN_KEYWORDS and TITLE_SCORES.

---

Running Tests

cd backend
node --test scorer.test.js   # scoring logic smoke tests
node --test scraper.test.js  # US-only filter + dedup tests

No test framework needed — uses Node's built-in node:test.

---

Architecture

backend/
  server.js        Express API (port 3001)
  scraper.js       Multi-source job fetcher + URL validator
  scorer.js        Hybrid scoring (deterministic + LLM)
  storage.js       Local JSON persistence + incremental scan cache
  config.js        All search queries and candidate profile
  utils/
    cache.js       In-memory scored job cache
    retry.js       Exponential backoff retry utility
    validator.js   Job schema validation
    env.validator  Startup env check

frontend/
  src/
    main.jsx       Vite entry point
    App.jsx        Main React app
    App.css        Dark-theme design system
  vite.config.js   Proxy to backend + .jsx resolution

• Backend: Node.js + Express (ESM)
• Frontend: React 18 + Vite (migrated from Create React App)
• AI Scoring: Groq LLaMA 3.1-8b-instant
• Storage: Local data.json (gitignored)
• LinkedIn: Apify Actor curious_coder~linkedin-jobs-scraper