CareerPilot AI – Intelligent Career & Placement Agent

Empowering Job Seekers with Local GPU-Accelerated Agentic AI

CareerPilot AI is a production-grade, full-stack autonomous career mentorship application built for the AMD AI DevMaster Hackathon (Agentic AI Track). It helps college students, fresh graduates, and professionals map skill gaps, audit resume ATS scores, practice speech-driven mock interviews, and organize custom 4-week study roadmaps.

---

🛠️ Tech Stack & Architecture

• Frontend: React (Vite), TypeScript, Tailwind CSS, Recharts (analytics dashboards), Framer Motion (glassmorphic layouts), Lucide Icons.
• Backend: Node.js, Express, TypeScript, REST APIs, MongoDB + Mongoose, JWT Authentication, Multer (file parser buffers), bcryptjs.
• AI Engine: Modular Dual-Mode Client supporting local Ollama (ROCm AMD accelerated) or cloud OpenAI API.

graph TD
    Client[React + Vite Frontend] -->|API Requests / JWT| Server[Express + TS Server]
    Server -->|Read/Write| DB[(MongoDB Database)]
    Server -->|Text Buffer| Parser[PDF Parse Engine]
    Server -->|AI Broker Broker| ClientAI{AI Config Layer}
    
    ClientAI -->|Local mode / ROCm| Ollama[Ollama Server http://localhost:11434]
    Ollama -->|GPU Accelerated| AMDGPU[AMD Radeon GPU via ROCm]
    
    ClientAI -->|Cloud mode| OpenAI[OpenAI API Cloud Endpoint]
    
    subgraph Local Inference AMD Environment
        Ollama
        AMDGPU
    end

Loading

---

🚀 AMD ROCm & GPU Acceleration Guide

CareerPilot AI features a modular inference broker specifically designed to route workloads to a local Ollama server running on AMD Radeon GPUs through ROCm.

1. Requirements

• AMD Radeon™ GPU (ROCm supported architectures: RDNA 2, RDNA 3, CDNA e.g., RX 6000, RX 7000 series, or workstation cards).
• Ollama installed locally on your Windows/Linux machine.
• ROCm compatible runtime drivers.

2. Configure AMD GPU Acceleration on Windows

For many consumer Radeon cards (like RX 6600, 6700, 7600 etc.), Ollama automatically detects the card. If it does not, you can enforce graphic library compatibility by setting Windows environment variables:

# In PowerShell (run as Administrator before starting Ollama)
$env:HSA_OVERRIDE_GFX_VERSION="10.3.0" 

# Or for RDNA 3 cards (e.g. RX 7000 Series)
$env:HSA_OVERRIDE_GFX_VERSION="11.0.0"

3. Start Local Inference Server

Pull and run the default lightweight models using Ollama:

# Pull the Llama 3 model (8 Billion Parameters)
ollama pull llama3

# Run the local server endpoint
ollama run llama3

Ollama will now host an OpenAI-compatible REST API at http://localhost:11434/v1 backed by AMD ROCm GPU acceleration.

4. Enable ROCm Mode in CareerPilot AI

In your backend/.env file, configure the AI Broker:

AI_PROVIDER=ollama
OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_MODEL=llama3

---

📁 Repository Directory Structure

e:/AMD/
├── backend/
│   ├── src/
│   │   ├── config/          # MongoDB Adapter, AI Client Configuration (ROCm vs Cloud)
│   │   ├── controllers/     # API logic (Auth, Chat, Resume, Interview, Roadmap, Dashboard)
│   │   ├── middleware/      # Auth tokens, Multer memory storage, Error handlers
│   │   ├── models/          # Mongoose Schemas (User, Chat, Resume, Interview, Roadmap, Notification)
│   │   ├── routes/          # Express route bindings
│   │   ├── services/        # AI Service prompt brokers and structured parser helpers
│   │   ├── utils/           # Helper scripts
│   │   └── index.ts         # Backend server bootloader
│   ├── package.json
│   ├── tsconfig.json
│   └── .env.example
├── frontend/
│   ├── public/              # Global assets, favicons, logos
│   ├── src/
│   │   ├── components/      # UI components (GlassCard, Layout, EmptyState indicators)
│   │   ├── context/         # Auth contexts, Themes modes caching, Toast notifications
│   │   ├── hooks/           # useSpeech (Mic inputs, Web Speech Synthesis output)
│   │   ├── pages/           # Pages (Dashboard, Resume, Interview, Roadmaps, Settings)
│   │   ├── App.tsx          # Central router guard
│   │   ├── index.css        # Tailwind + Glassmorphism directives
│   │   └── main.tsx
│   ├── package.json
│   ├── tsconfig.json
│   ├── tailwind.config.js
│   └── vite.config.ts

---

🔧 Installation & Local Launch

Ensure you have Node.js (v18+) and MongoDB installed and running locally.

Step 1: Set Up Backend Configurations

cd backend
npm install

Rename the .env.example file to .env and adjust variables if needed:

PORT=5000
MONGODB_URI=mongodb://localhost:27017/careerpilot
AI_PROVIDER=ollama
OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_MODEL=llama3

Step 2: Set Up Frontend Configurations

cd ../frontend
npm install

Step 3: Run the Development Servers

You can launch both applications in separate terminals:

# In terminal 1 (start backend API on port 5000)
cd backend
npm run dev

# In terminal 2 (start Vite React frontend on port 3000)
cd frontend
npm run dev

Open your browser and navigate to: http://localhost:3000

---

👥 Seed Demo Accounts

To test the application immediately without registration, you can use the default test account credentials:

• Email: testpilot@careerpilot.ai
• Password: password123

(These test profiles will automatically hook up default dashboard metrics, resume reviews history logs, and learning roadmaps, enabling rapid testing).

---

📡 REST API Documentation Reference

All requests must include the JWT token header: Authorization: Bearer <token> (except public auth endpoints).

Method	Endpoint	Description	Payload
POST	/api/auth/signup	Register new profile account	{ name, email, password }
POST	/api/auth/login	Login and return JWT token	{ email, password }
GET	/api/auth/profile	Retrieve active profile details	None
PUT	/api/auth/profile	Update profile / target career	{ name, targetCareer }
GET	/api/dashboard/overview	Retrieve dashboard stats & checklist	None
POST	/api/resume/analyze	Parse PDF resume and calculate score	file (FormData), targetJob
GET	/api/resume/latest	Retrieve latest resume ATS critique	None
POST	/api/interview/start	Start new mock interview session	{ type: 'technical'|'hr', role }
POST	/api/interview/:id/answer	Submit answer and get next question	{ answer }
POST	/api/roadmap/analyze-gap	Map skill gap & generate 4-week roadmap	{ dreamJob, currentSkills }
PUT	/api/roadmap/toggle-task	Toggle checklist task completion state	{ taskId, isCompleted }
GET	/api/chat	Get conversation chats list	None
POST	/api/chat	Create new consultation chat	{ title }
POST	/api/chat/:id/message	Send message to AI mentor	{ content }

---

🏆 Gamified Achievement Badges

CareerPilot AI builds user engagement using unlockable achievements tracked on the profile dashboard page:

1. Welcome Pilot: Account created successfully.
2. Profile Completed: Career objectives and current skills populated.
3. Resume Critic: Uploaded first PDF resume.
4. ATS Master: Achieved a resume ATS rating of 80% or higher.
5. Interview Explorer: Completed first mock interview loop.
6. Mock Superstar: Achieved a mock score rating of 8.0/10 or higher.
7. Roadmap Planner: Generated first week-by-week learning plan.
8. Pathfinder: Checked off first learning task.
9. Placement Graduate: Achieved 100% completion on a roadmap.
10. Deep Thinker: Maintained 5 or more consultations chats.