Job Hunter SG

▶ 30-second tour · Live app · kooexperience.com
_{Music: Sascha Ende, ende.app (CC BY 4.0)}

AI-powered job aggregator, resume coach, and career toolkit for the Singapore job market.

Searches MyCareersFuture and Careers@Gov in one interface. Scores your resume, rewrites bullets with validation gates to prevent hallucination, and generates tailored cover letters. Semantic matching via RAG finds jobs keyword search would miss.

Try it live | Blog Post | Portfolio

See CHANGELOG.md for recent enhancements and fixes.
Maintainers should start with the authoritative maintainer handbook.

Screenshots

Job Search — MCF + Careers@Gov with filters	Upload — PDF, paste, AI chat, or demo
Templates — 8 professional styles	AI Feedback — per-bullet coaching
AI Rewrite — 3 options per bullet	Market Insights — sector & skill trends
Smart Match — RAG-powered job matching	Export — DOCX with final score

Search Visibility

• Public app URL: https://job.kooexperience.com/
• robots.txt allows search engines plus OpenAI search crawlers for the public app while blocking /api/ and /api/admin/.
• sitemap.xml lists the canonical app URL.
• llms.txt gives AI assistants a concise machine-readable project summary.
• index.html includes canonical, Open Graph, Twitter Card, WebApplication JSON-LD, and FAQ JSON-LD metadata.

After deployment, submit the site to Google Search Console and Bing Webmaster Tools, then confirm https://job.kooexperience.com/robots.txt, /sitemap.xml, and /llms.txt are publicly reachable.

Features

Job Search

• Aggregates MyCareersFuture and Careers@Gov; current source counts and freshness are shown in the dashboard — CareersGov data via OpenGovSG (credit: Alwyn Tan @ OGP)
• Production source integrations should use official APIs, public employer feeds, or documented ATS job-board endpoints; browser automation is kept out of the hosted crawler unless explicitly reviewed for legal, privacy, and reliability risk
• Nightly crawl via Railway cron (22:00 UTC) for MyCareersFuture and Careers@Gov; Adzuna and Jooble are optional API integrations
• Filter by seniority, job type, salary range, skills
• ATS skill tags extracted at scrape time using the maintained skill_extractor taxonomy
• Source-aware dedupe uses official posting IDs or canonical source URLs, so repeated keyword hits collapse without merging distinct postings from the same employer
• Precomputed sector, salary floor, official SSIC fields, and skill-search fields keep filters fast without loading full job tables

Resume Builder

• Upload PDF/DOCX or build from scratch with AI chat; signed-in uploads are saved as distinct resume versions
• 8 professional templates (Classic, Modern, SG Professional, Compact, Executive, Creative, Technical, Minimal)
• Inline click-to-edit with drag-and-drop bullet reordering (Framer Motion)
• Add/delete sections, entries, bullets with one click
• Download as DOCX via python-docx

AI Resume Coach (SEA-LION)

• ATS scoring (0-100) across Impact, Presentation, Competencies
• Per-bullet feedback with annotations (Solid Impact, Review, Verb Check)
• AI rewrite with 3 validated options per bullet
• 7-stage tailoring pipeline for specific JDs; active model tiers and budgets are configured in backend/config.py
• Deterministic validation gates check fact preservation, AI phrasing, keywords, length, and hallucination risk on every rewrite
• Injectable vs non-injectable keyword classification — AI can only add skills the user plausibly has
• Custom summary generation with user prompts

AI Resume Chat Builder

• Conversational interface that builds your resume from scratch
• Coaches you to add metrics and quantified achievements
• Generates structured resume dropped directly into the editor

Cover Letter Generator

• Generate from resume + job description
• Custom direction ("emphasize leadership", "keep it concise")
• Edit inline, copy, or download

Smart Match (RAG)

• Semantic job matching using sentence-transformers/all-MiniLM-L6-v2 (384-dim)
• Hybrid scoring: keyword overlap + cosine similarity
• Suitability scores, skill gap analysis, bridge paths
• Shows the exact stored resume snapshot used for matching so results are auditable
• Pre-embedded jobs for instant matching
• Persisted match snapshots return repeat visits instantly when resume and job corpus are unchanged
• "Close the Gap" course recommendations map repeated missing skills to official MySkillsFuture courses, ranked by relevance, rating, career impact, and response count

Market Insights

• Industry/sector, title, company, salary, freshness, and seniority breakdowns across the cached job corpus
• Official company industry uses ACRA corporate entity data from data.gov.sg when an employer maps to SSIC; otherwise the app keeps the sector explicitly labelled as inferred or unavailable
• Drill down by source, industry/sector, title, or company/department to inspect Careers@Gov, MyCareersFuture, civil-service demand, and hiring concentration
• Directional market movers compare last-30-day dated postings against older postings in the same slice; persisted daily snapshots are the next step for true trend charts
• Salary views show advertised floor plus midpoint when range data is available
• Sector inference uses precomputed title and extracted-skill signals only as the fallback when official ACRA/SSIC mapping is unavailable

Application Tracker

• Track applications: Applied, Interview, Offer
• Follow-up reminders
• Status tracking per job

Interview Story Bank

• Generates STAR+R interview story drafts from resume evidence with review-before-save flow
• Manual Big Three prompts cover elevator pitch, impact project, and conflict resolution
• Long-running story extraction shows live progress feedback so users know it is still working

Tech Stack

Quick Start

The authoritative fresh-clone instructions, prerequisites, and environment
boundaries are in docs/getting-started.md. The short
development loop is:

# Backend
python3.12 -m venv .venv
.venv/bin/python -m pip install -r backend/requirements.txt
.venv/bin/python backend/main.py  # starts on :8000

# Frontend
cd frontend
npm ci --legacy-peer-deps
npm run dev     # starts on :5173, proxies /api to :8000

Testing

# Fast backend quality checks
python -m compileall -q backend
ruff check backend tests

# Frontend production build
cd frontend && npm run build

# Full test suites
PYTHONPATH=backend python -m pytest backend/tests tests -q
cd frontend && npm test

Local pre-commit hooks are intentionally lightweight and CI is the merge gate:

pip install pre-commit
pre-commit install

Architecture

See docs/architecture.md for the maintained end-to-end
flow and trust boundaries. This compact map is only an orientation:

configured sources → sanitize/precompute → scraped_jobs → API ↔ React
guest resume upload → isolated parser → canonical browser document
signed-in resume upload → isolated parser → immutable saved resume version
saved resume versions → scoring / Power Match / Documents
classic tailoring or accepted recruitment edits → new derived resume version
tracked_jobs.role_metadata → latest application cover letters → Documents

API Highlights

60+ endpoints total. Key ones:

Environment Variables

# Omit DATABASE_URL locally to use backend/jobhunter.db.
# Set an explicit PostgreSQL URL on hosted deployments.
JWT_SECRET=your-secret
AUTH_MODE=password
ACCOUNT_AI_PER_DAY=500
SEALION_API_KEYS=your-key,another-key # canonical comma-separated key pool
ALLOWED_EMAIL_DOMAINS=*               # or comma-separated domains
ALLOWED_ORIGINS=http://localhost:5173  # CORS

# ACRA SSIC company taxonomy
# Default: local backend/data/company_ssic_cache.json only; no live lookup on user requests.
# Local backfill:
#   cd backend && python backfill_company_ssic.py --limit 200 --live --delay 8
# Railway production backfill:
#   railway ssh python backfill_company_ssic.py --limit 200 --live --delay 8
# Do not use `railway run` for this; it runs locally and Railway's private
# postgres.railway.internal hostname will not resolve from your laptop.
# COMPANY_SSIC_CACHE_PATH=/absolute/path/company_ssic_cache.json
ACRA_LIVE_LOOKUP=0

See .env.example for the maintained operational reference and
backend/config.py for lower-level tuning defaults.

License

AGPL-3.0 — Free to use and modify. Must share changes if deployed as a service. Attribution required.

Built by Haoming Koo.