CoolFace
Apppublic

lablab-ai-amd-developer-hackathon/pakfreelance-ai-agent

sourceHugging Faceupdated 5mo agoView on Hugging Face
1likes
App README

🇵🇰 PakFreelance — AI Freelance Toolkit

Professional AI Agent Toolkit for Pakistani Freelancers AMD Developer Hackathon 2026 | Track 1: AI Agents & Agentic Workflows

Powered by AMD MI300X through Groq's optimized API, running Llama 3.3 70B for fast inference. Deployed on HuggingFace Spaces.

🎯 Overview

PakFreelance is an AI-powered toolkit built with CrewAI agents that helps freelancers across Pakistan and South Asia:

✅ Generate winning proposals in seconds (5-agent AI orchestration) ✅ Detect client scams before wasting time (red flag analysis) ✅ Calculate optimal rates based on skills and goals ✅ Score proposals with detailed improvement feedback ✅ Write professional bios for Upwork/Fiverr ✅ Analyze client websites for legitimacy ✅ One-click Power Mode for complete job analysis

Built for the 4M+ freelancers in Pakistan and South Asia using AMD MI300X infrastructure, Llama 3.3 70B, and CrewAI.

🚀 Key Features

⚡ Power Mode (One-Click Analysis)

Paste a job posting → Get risk assessment + proposal + pricing strategy in 3 columns

✈️ Proposal Generator

5 specialized AI agents working together:

  1. 1.Scout - Extract job requirements
  2. 2.Researcher - Analyze client context
  3. 3.Writer - Generate personalized proposal (150-200 words)
  4. 4.Pricer - Recommend optimal pricing
  5. 5.Reviewer - Polish & finalize

🚩 Red Flag Detector

  • —Identifies scam patterns
  • —Detects unrealistic budgets
  • —Spots exposure-only offers
  • —Provides safe response templates

⭐ Proposal Scorer

  • —Scores out of 100
  • —Breakdown by 7 criteria (Hook, Credibility, Understanding, Solution, etc.)
  • —Top 3 specific improvement suggestions

💰 Rate Calculator

  • —Pakistan-specific pricing (30-50% lower than US rates)
  • —Experience multiplier
  • —Monthly goal calculation
  • —Tax & platform fee buffer
  • —Three-tier pricing strategy (Start/Target/Premium)

📝 Profile Bio Writer

  • —Upwork/Fiverr optimized
  • —Compelling headlines
  • —Professional 150-200 word bios
  • —Personality-driven tone

🌐 Website Analyzer

  • —Extract website content
  • —Analyze for red flags
  • —Assess client legitimacy

📄 Document Reader

  • —Upload CV & Job PDFs
  • —Extract relevant info
  • —Provide smarter analysis

🛠️ Tech Stack

ComponentTechnologyPurpose
AI FrameworkCrewAIMulti-agent orchestration
LLMLlama 3.3 70B (via Groq)Fast inference on AMD MI300X
InfrastructureAMD MI300X (via Groq API)GPU acceleration
Web FrameworkStreamlitBeautiful UI
Document Exportpython-docxWord document generation
PDF SupportPyPDF2PDF reading
Web ScrapingrequestsWebsite text extraction

📋 System Architecture

┌─────────────────────────────────────────────────────────┐
│                  PAKFREELANCE APP                       │
│              (Streamlit Web Interface)                  │
└──────────────────┬──────────────────────────────────────┘
                   │
        ┌──────────┴──────────┐
        │                     │
        ▼                     ▼
    ┌────────────┐      ┌──────────────┐
    │  Job Input │      │ User Profile │
    └──────┬─────┘      └──────┬───────┘
           │                   │
           └───────┬───────────┘
                   ▼
        ┌──────────────────────┐
        │   CrewAI Agents:     │
        ├──────────────────────┤
        │ • Scout Agent        │
        │ • Researcher Agent   │
        │ • Writer Agent       │
        │ • Pricer Agent       │
        │ • Reviewer Agent     │
        └───────┬──────────────┘
                │
                ▼
        ┌──────────────────────┐
        │  Groq API Call       │
        │ (Llama 3.3 70B)      │
        │                      │
        │ AMD MI300X Powered   │
        └───────┬──────────────┘
                │
                ▼
        ┌──────────────────────┐
        │  Output Results:     │
        ├──────────────────────┤
        │ • Proposals          │
        │ • Red Flags          │
        │ • Rates              │
        │ • Scores             │
        │ • Bios               │
        └──────────────────────┘

📦 Installation

1. Clone Repository

bash
git clone https://github.com/yourusername/pakfreelance.git
cd pakfreelance

2. Create Virtual Environment (Optional but Recommended)

bash
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install Dependencies

bash
pip install -r requirements.txt --break-system-packages

4. Set Up Environment Variables

Create .env file in project root:

env
GROQ_API_KEY=your_groq_api_key_here

Get API Keys:

  • —Groq: https://console.groq.com (Free tier: 12,000 TPM)

5. Run Application

bash
streamlit run app.py

Visit: http://localhost:8501

🎮 Usage Guide

Power Mode (Quickest)

  1. 1.Click ⚡ Power Mode in sidebar
  2. 2.Paste job posting
  3. 3.Click 🚀 Run Complete Analysis
  4. 4.Get risk assessment + proposal + rates in 3 columns
  5. 5.Click expanders for full results

Proposal Generator

  1. 1.Paste job posting
  2. 2.(Optional) Enter your skill & experience
  3. 3.Click ✈️ Generate Proposal
  4. 4.Review proposal
  5. 5.Download as Word document (.docx)

Red Flag Detector

  1. 1.Choose: Job Description or Website URL
  2. 2.Paste job or enter website
  3. 3.Click 🔍 Analyze for Red Flags
  4. 4.Get risk level + specific red flags
  5. 5.Copy safe response template if HIGH RISK

Rate Calculator

  1. 1.Enter skill (e.g., "Python Developer")
  2. 2.Enter years of experience
  3. 3.Enter monthly income goal (USD)
  4. 4.Enter available hours/week
  5. 5.Click 💰 Calculate Rate
  6. 6.Get optimal hourly rate + pricing strategy

Proposal Scorer

  1. 1.Paste your proposal
  2. 2.Click ⭐ Score My Proposal
  3. 3.Get score out of 100 + breakdown + improvements

Profile Bio Writer

  1. 1.Enter name, skill, experience
  2. 2.(Optional) Specialization & achievements
  3. 3.Click 📝 Generate Bio
  4. 4.Get headline + professional bio

📊 Example Output

Proposal Generated

Professional Proposal
Generated by: PakFreelance AI Toolkit

Building a complete e-commerce platform with a MERN stack is exactly 
the kind of challenge I enjoy solving. I have 5+ years of experience 
building scalable applications for e-commerce clients, including 10+ 
projects with React, Node.js, and MongoDB.

I'll design a responsive UI using React, implement a robust backend 
with Node.js and MongoDB, integrate Stripe for payments, and set up 
JWT authentication. You'll get a GitHub repo and full documentation.

I'm available immediately and can deliver within your 20-day timeline...

Best regards,
[Your Name]

---
Generated by PakFreelance AI
Powered by AMD MI300X • CrewAI • Llama 3.3 70B

Rate Calculation

OPTIMAL HOURLY RATE: $15 - $30/hour

CALCULATION BREAKDOWN:
- Monthly goal: $3000
- Weeks per month: 4.3
- Hours per week: 30
- Base rate needed: $22.08/hour
- Market rate for Prompt Engineering: $25 - $50/hour
- Experience multiplier (2 years): 1.2
- Adjusted rate: $20 - $40/hour
- Considering Pakistani context: $15 - $30/hour

PRICING STRATEGY:
- Start at: $15/hour (to build reviews)
- Target rate: $25/hour (once established)
- Premium rate: $35/hour (for repeat clients)

🏗️ Project Structure

pakfreelance/
├── app.py                    # Main Streamlit app (650+ lines)
├── agents.py                 # CrewAI agents + tasks (150+ lines)
├── red_flags.py              # Red flag detection agent
├── proposal_scorer.py        # Proposal scoring agent
├── rate_calculator.py        # Rate calculation agent
├── profile_bio_writer.py     # Bio generation agent
├── requirements.txt          # Python dependencies
├── .env.example              # Example environment variables
├── README.md                 # This file
└── .gitignore               # Git ignore patterns

📋 Dependencies

crewai==0.25.0              # Multi-agent orchestration
streamlit==1.28.0           # Web UI framework
groq==0.4.1                 # Groq API client
litellm==1.0.0              # LLM wrapper
python-dotenv==1.0.0        # Environment variables
python-docx==0.8.11         # Word document creation
PyPDF2==3.0.1               # PDF reading
requests==2.31.0            # HTTP requests
duckduckgo-search==3.8.4    # Web search (optional)

Full list: See requirements.txt

🎯 AMD Technology Integration

Infrastructure Overview

This project leverages AMD MI300X GPUs through Groq's optimized API:

  • —LLM: Llama 3.3 70B (AMD-optimized model)
  • —Inference: Groq API (runs on AMD MI300X)
  • —Performance: Fast token generation (100+ tokens/sec)
  • —Hosting: HuggingFace Spaces (production-ready)

How AMD Powers This Application

  1. 1.AMD MI300X GPU Compute
  2. 2.All AI inference runs on AMD Instinct MI300X architecture
  3. 3.Powered via Groq's AMD-optimized endpoints
  4. 4.Fast token generation (100+ tokens/sec)
  5. 5.Cost-effective for production workloads
  1. 1.ROCm Support
  2. 2.Llama 3.3 70B is ROCm-compatible
  3. 3.Optimized for AMD Instinct MI300X
  4. 4.Can scale to local AMD GPU clusters with ROCm
  1. 1.CrewAI Multi-Agent Orchestration on AMD
  2. 2.5 specialized agents working in sequence
  3. 3.All processing runs on AMD MI300X hardware
  4. 4.Demonstrates scalable agent architecture
  1. 1.Why AMD for This Project
  2. 2.✅ Cost-effective inference (free API tier)
  3. 3.✅ Production-ready infrastructure
  4. 4.✅ Supports open-source models (Llama 3.3)
  5. 5.✅ Scalable to enterprise deployments
  6. 6.✅ No vendor lock-in (ROCm is open source)

🚀 Deployment

Hugging Face Spaces (Recommended)

  1. 1.Create Hugging Face Space at https://huggingface.co/spaces
  2. 2.Select Streamlit SDK
  3. 3.Connect GitHub repository
  4. 4.Add environment variable: GROQ_API_KEY
  5. 5.Deploy automatically (3-5 minutes)
  6. 6.Get live URL for submission

Local Testing

bash
streamlit run app.py

🐛 Troubleshooting

"RateLimitError from Groq"

Wait 8 seconds between requests. Free tier: 12,000 TPM (tokens per minute)

"Invalid API Key"

Check .env file - make sure key starts with gsk_ and has no extra spaces

"ModuleNotFoundError"

bash
pip install module-name --break-system-packages

📝 License

MIT License - See LICENSE file for details

🙏 Acknowledgments

  • —AMD Developer Program - Infrastructure & credits
  • —lablab.ai - Hackathon platform
  • —CrewAI - Multi-agent orchestration framework
  • —Groq - Fast LLM inference on AMD MI300X
  • —Llama 3.3 70B - Powerful open-source language model
  • —Streamlit - Beautiful web framework

📞 Support

  • —Documentation: See this README
  • —Issues: Open a GitHub issue
  • —Setup Help: Check .env.example for environment variables

🚀 Future Roadmap

Current (v1.0)

  • —✅ Multi-agent proposal generation
  • —✅ Red flag detection
  • —✅ Rate calculation
  • —✅ Proposal scoring
  • —✅ Bio generation
  • —✅ Website analysis
  • —✅ Power Mode

Planned (v2.0)

  • —Document upload & analysis
  • —Browser extension
  • —Mobile app
  • —Multi-language support

Stretch Goals (v3.0)

  • —Fine-tuned models
  • —Real-time job scraping
  • —Auto-apply features
  • —Client management dashboard

Built with ❤️ for Pakistani Freelancers Powered by AMD MI300X through Groq AMD Developer Hackathon 2026 - Track 1: AI Agents

Last Updated: May 8, 2026