Kenfack-franck

Posted on Feb 5

How I Built a Full-Stack AI Job & Internship Search Platform in 6 Days with GitHub Copilot CLI

#github #githubcopilot #webdev #ai

This is a submission for the GitHub Copilot CLI Challenge

What I Built

JobHunter AI - An AI-powered platform that automates job AND internship searches for professionals and students.

The Problem I Solved

Whether you're a professional looking for a job or a student searching for an internship, the process is exhausting:

Browse Indeed → LinkedIn → RemoteOK → WTTJ... hours wasted
Customize your resume/CV for each application manually
Lose track of where you applied and when
No idea if you're qualified or wasting time applying

For students: Finding internships is even harder - most platforms focus on full-time jobs, not internships or "stages".

On January 29, 2026, I thought: "What if AI could handle all of this for BOTH job seekers and students?"

The Solution

A complete platform that serves both professionals and students:

✅ Multi-Source Search - Searches 6+ platforms simultaneously for jobs AND internships (Indeed, LinkedIn, RemoteOK, WTTJ, Adzuna, TheMuse)

✅ AI Compatibility Scoring - Uses pgvector + OpenAI embeddings to score each opportunity 0-100%

✅ Smart CV Parsing - Upload your PDF resume, AI extracts experiences, education, skills automatically

✅ AI Document Generation - Personalized resumes and cover letters with GPT-4 + Gemini in 30 seconds

✅ Application Tracking - Manage all your job/internship applications in one dashboard

✅ Fully Responsive - Perfect mobile experience for job hunting on-the-go

The Challenge

Build all of this in less than a week.

Spoiler: With GitHub Copilot CLI, I did it in 6 days.

Tech Stack

Backend: FastAPI, PostgreSQL + pgvector, Redis + Celery, OpenAI GPT-4, Google Gemini

Frontend: Next.js 14, TypeScript, TailwindCSS, shadcn/ui

DevOps: Docker Compose, Caddy reverse proxy, VPS

Stats: 68,214 lines added, 85 commits, 320 files changed, 6 days from idea to production

Demo

🌐 Live Application

URL: https://jobhunter.franckkenfack.works

GitHub Repository: https://github.com/Kenfack-franck/jobHunterAi

video: https://drive.google.com/file/d/1J0qqiPMVR7w7Oo6hbdgwP2S_QtJGNvs-/view?usp=drive_link

🔐 Test Credentials

To test the full experience without creating an account:

Role	Email	Password
User	test@test.com	Test123!

✨ What You Can Test

📄 Upload Your CV - Watch AI parse your resume automatically
🔍 Search Jobs/Internships - Try "Python Developer", "Data Analyst Internship", or "Marketing Stage"
🎯 View Compatibility Scores - See detailed AI analysis of why you match (or don't) each opportunity
📝 Generate Documents - Get a personalized resume tailored to a specific job in 30 seconds
📊 Track Applications - Manage all your job and internship applications
📱 Mobile Experience - Open on your phone, it's fully responsive

📸 Application Screenshots

Landing page with clear value proposition for both professionals and students

)
Simple registration flow

)
Secure authentication system

)
Interactive onboarding guide for new users

)
Main dashboard with real-time search progress and stats

)
Multi-step profile creation: Personal information

)
Profile creation: Skills and experiences

)
Profile creation: Job preferences and expectations

)
Choose which job boards to search (Indeed, LinkedIn, RemoteOK, etc.)

)
Real-time multi-source job/internship search with progress tracking

)
Advanced filters: location, contract type, remote work, salary

)
Search results with AI compatibility scores (0-100%)

)
Detailed job description with compatibility analysis

)
Saved opportunities organized by compatibility score

)
Generate personalized CV and cover letter in 30 seconds

)
Track all applications with status updates and notes

)
User settings and preferences

)
Contact page for user support

)
Comprehensive help documentation

)
Fully responsive mobile interface for job hunting anywhere

My Experience with GitHub Copilot CLI

Let me be brutally honest: Without GitHub Copilot CLI, this would have taken 3-4 weeks. With it, I shipped in 6 days.

Here's exactly how I used it and the real, measurable impact.

📊 The Real Numbers (100% Verifiable!)

I documented everything meticulously. Here are the actual statistics from my Copilot CLI usage logs:

Metric	Value	Source
Duration	6 days (Jan 30 - Feb 4)	Git commit history
Copilot Sessions	5 major sessions	Session state logs
My Messages to Copilot	340+ messages	events.jsonl
Copilot Responses	4,057 responses	Measured from logs
Tool Executions	4,988 commands	bash/view/edit/grep calls
Git Commits	85 commits	`git log --oneline`
Files Changed	320 files	`git diff --stat`
Lines Added	68,214 lines	`git log --stat`
Lines Deleted	32,170 lines	Refactoring + cleanup
Feature Commits	38 commits	Keyword analysis
Bug Fixes	43 commits	Fix/debug commits

All evidence is public: Check the copilot-evidence/ folder in my GitHub repo for raw data, logs, and session transcripts.

💡 9 Real Examples With Actual Commands

These aren't hypothetical scenarios. These are actual problems I encountered with the exact Copilot CLI commands I used and the real impact.

1️⃣ Docker Build Failing at 3 AM (Feb 1)

Problem: Production Docker build failing with mysterious module errors:

Module not found: Can't resolve '@/lib/applicationsService'
Module not found: Can't resolve '@/lib/auth'
Module not found: Can't resolve '@/components/JobCard'

It's 3 AM. I'm exhausted. The modules exist locally. Spent 30 minutes debugging.

Command I Used:

@github-copilot explain why Docker build can't find modules that exist locally but build works fine in development

Copilot's Response:

"This is a Docker context issue. Your files aren't being copied to the build stage before npm run build runs. You need to add COPY . . before the build command in your Dockerfile."

Result: Added one line to Dockerfile. Build succeeded. Deployed to production.

Time Saved: 1-2 hours of frustration ⏱️

2️⃣ TypeScript Type Mismatch Hell (Feb 1)

Error Flooding Console:

Argument of type 'number' is not assignable to parameter of type 'string'
Property 'work_mode' does not exist on type 'JobOffer'
Type 'undefined' is not assignable to type 'string'

8 different type errors across multiple files.

Command:

@github-copilot check why docId is typed as number when Document.id should be string in the schema

Copilot: Showed me the schema mismatch and suggested the exact type definitions to fix.

Result: Fixed all 8 errors in under an hour.

Time Saved: 2-3 hours 📝

3️⃣ The CORS Mystery That Wasn't CORS (Feb 3)

Problem: Production completely broken. CORS errors everywhere:

Access to XMLHttpRequest blocked by CORS policy
No 'Access-Control-Allow-Origin' header present

But I HAD configured CORS middleware correctly! Spent 30 minutes checking CORS config.

Command:

@github-copilot explain CORS error even though CORS middleware is properly configured in FastAPI

Copilot's Response:

"CORS errors can be misleading. They often appear when the API server crashes before sending headers. Check if your backend is actually running and check the server logs for the real error."

Checked logs. Real issue: Database migrations hadn't been applied! The user.role column didn't exist.

Copilot saved me from: Debugging the wrong thing for hours.

Time Saved: 2+ hours 🎯

4️⃣ Database Migration Crisis (Feb 3)

Production Error:

sqlalchemy.exc.ProgrammingError: column users.role does not exist

Command:

@github-copilot suggest exact command to apply pending Alembic migrations in production Docker container

Copilot:

docker compose -f docker-compose.prod.yml exec backend alembic upgrade head

Ran it. Migrations applied. Production fixed.

Time Saved: 20-30 minutes (no need to look up Alembic docs)

5️⃣ Mobile Responsiveness Nightmare (Feb 3)

Problem: Job search modal was completely unusable on mobile devices. Too wide, cut off, unscrollable.

Command:

@github-copilot suggest how to make this TailwindCSS modal responsive for mobile screens

Copilot's Suggestions:

Change md:grid-cols-3 to sm:grid-cols-3
Add grid-cols-1 for mobile-first
Use p-4 sm:p-8 for responsive padding
Add max-h-screen overflow-y-auto

Result: Changed 10 lines. Perfect mobile experience.

Time Saved: 1 hour of trial-and-error 📱

6️⃣ Async Job Search Architecture (Jan 31)

Problem: Job search was blocking the API for 30-60 seconds while scraping 6 different job boards. Terrible UX - users thought the app was frozen.

Command:

@github-copilot suggest FastAPI endpoint architecture for async job search with real-time progress updates

Copilot's Complete Architecture:

Create /jobs/search/async endpoint that returns task ID immediately
Use Celery to run scraping in background
Create /jobs/search/status/{task_id} endpoint for polling
Return progress states: pending, processing, completed, failed
Frontend polls every 2 seconds for updates

Implementation: Built the entire async system in 2 hours following this architecture.

Result: Instant response, smooth progress bar, perfect UX.

Time Saved: 3-4 hours of research and trial-and-error 🚀

7️⃣ Git Repository Cleanup (Feb 4)

Problem: My repo had 60+ markdown documentation files polluting the root directory. Looked unprofessional.

Command:

@github-copilot suggest .gitignore pattern to exclude all .md files except README.md

Pattern:

*.md
!README.md

Follow-up Command:

git ls-files "*.md" | grep -v "README.md" | xargs git rm --cached

Result: Clean repository structure in 5 minutes.

Time Saved: 10-15 minutes 🧹

8️⃣ Missing TypeScript Properties (Feb 1)

Error:

Property 'work_mode' does not exist on type 'JobOffer'
Property 'salary_max' does not exist on type 'JobOffer'

Command:

@github-copilot check if work_mode and salary_max exist in backend JobOffer model

Copilot: Found them in backend, missing in frontend TypeScript interface.

Fix: Added 2 lines to interface. Done.

Time Saved: 10 minutes

9️⃣ Admin Panel Mobile Navigation (Feb 4, Today!)

Problem: Admin users couldn't access admin panel on mobile - link was missing from mobile navigation menu.

Command:

@github-copilot add admin dashboard link to mobile navigation menu based on user role

Copilot's Solution:

{user?.role === 'admin' && (
  <Link href="/admin" className="flex items-center gap-2">
    <Shield className="h-5 w-5" />
    <span>Admin</span>
  </Link>
)}

Result: Admin link now shows only for admin users on mobile.

Time Saved: 15-20 minutes

⏱️ Time Comparison: Before vs After

Task Category	With Copilot CLI	Without Copilot CLI
Docker debugging	1-2 hours	4-6 hours
TypeScript type errors (8 errors)	1 hour	3-4 hours
CORS diagnosis & fix	30 minutes	2-3 hours
Database migrations	20 minutes	1-2 hours
Responsive UI fixes	1 hour	3-4 hours
Async architecture design	2 hours	6-8 hours
Git cleanup	15 minutes	1 hour
Code refactoring	2-3 hours	8-10 hours
TOTAL PROJECT	6 days	3-4 weeks

Estimated Time Saved: 60-80 hours ⚡

🎓 Key Learnings From Using Copilot CLI

1. Pair Programming With a Senior Developer

Using Copilot CLI felt like having a senior developer sitting next to me:

"Why is this failing?" → Instant, context-aware answer
"How do I configure this?" → Exact command, no docs diving
"What's best practice here?" → Industry standards explained

The difference? This "senior dev" never gets tired, annoyed, or needs coffee breaks.

2. Debugging Became 10x Faster

Before Copilot CLI:

Google the error message
Stack Overflow (top 5 answers)
Try solution 1 (doesn't work)
Try solution 2 (doesn't work)
Page 3 of Google (find the real answer)
Understand, adapt, implement

With Copilot CLI:

Ask Copilot the error
Get context-specific answer
Implement
Fixed

Hours became minutes.

3. Learning Best Practices Without Reading Docs

Copilot naturally suggested:

Proper Alembic migration commands
Correct TailwindCSS responsive breakpoints
Git patterns I didn't know existed
Docker multi-stage build optimizations
FastAPI async best practices

I learned while building, not by reading documentation first.

4. Context Awareness is Game-Changing

Copilot CLI understood:

My FastAPI + Next.js stack
My shadcn/ui component library
My Docker Compose setup
My specific file structure
My database schema

Suggestions weren't generic - they were tailored to my exact project.

5. It's an Accelerator, Not Magic

Important reality check: Copilot CLI doesn't write entire features for you. But it:

Removes friction - No more "How do I...?" roadblocks
Saves research time - No documentation diving for syntax
Catches errors early - Before you waste time debugging
Points you in the right direction - Architecture suggestions
Handles boilerplate - Configuration files, type definitions

You still design the architecture and write the logic. But you build much, much faster.

🚧 Honest Challenges & Limitations

Being completely transparent:

Suggestions sometimes need adjustment - Copilot gives you the right direction, but you still need to adapt to your specific needs (~10-15% of the time)
You must understand the code - Don't blindly copy suggestions. Read, understand, then implement.
Not a replacement for thinking - You still design the architecture, plan features, make decisions. Copilot executes your vision faster.
Context limits matter - Very large files or complex multi-file changes require breaking down the problem.
10-15% suggestions aren't relevant - But 85% success rate is still incredible.

Bottom line: Copilot CLI is a superpower, but you're still the superhero.

🎯 The Bottom Line

Six days ago: An idea to help professionals and students find jobs/internships faster.

Today:

✅ Production application running on a VPS
✅ 15,000+ lines of TypeScript/Python code
✅ 40+ React components
✅ 28 API endpoints
✅ AI-powered features (embeddings, GPT-4, Gemini)
✅ Fully responsive design
✅ Real users finding real opportunities

Could I have built this without Copilot CLI?

Yes, absolutely.

Would it have taken 3-4 weeks instead of 6 days?

100% yes.

More importantly: Copilot CLI didn't just save time - it made development enjoyable. No frustration. No getting stuck. Just flow state.

Try JobHunter AI Today! 🚀

Live Application: https://jobhunter.franckkenfack.works

Test Credentials:

Email: test@test.com
Password: Test123!

GitHub Repository: github.com/Kenfack-franck/jobHunterAi

Perfect for:

💼 Professionals searching for their next job
🎓 Students looking for internships or "stages"
🚀 Anyone tired of manual job hunting across multiple sites
🤖 Developers interested in AI-powered applications

Let's Connect! 💬

Have questions? Want to discuss the technical implementation? Found a bug?

Drop a comment below! I'm happy to:

Explain any technical decisions
Share more detailed Copilot CLI examples
Discuss the AI/ML implementation
Talk about the async architecture

If you found this helpful:

⭐ Star the GitHub repo
🔄 Share this article with fellow developers
💬 Leave your feedback below
🐦 Connect with me on social media

Built with passion 🔥, powered by AI 🤖, accelerated by GitHub Copilot CLI ⚡

All statistics are 100% verifiable in the copilot-evidence/ folder of the GitHub repository.

P.S. I even used Copilot CLI to help structure and write parts of this article. Meta! 😄

Technical Implementation Highlights

For the technically curious, here are some interesting implementation details:

AI Compatibility Scoring

Uses OpenAI text-embedding-3-small to vectorize job descriptions and user profiles
Stores vectors in PostgreSQL with pgvector extension
Calculates cosine similarity for matching (0-100% score)
Caches results in Redis for performance