ML Engineer Portfolio & AI Demo Stack

Welcome to my comprehensive portfolio website featuring an interactive machine learning model collection and professional showcase.

🎯 About This Project

This is a full-stack React application that serves as both my personal portfolio and a demonstration platform for various machine learning and AI capabilities. The site showcases my work as a Machine Learning Engineer while providing interactive demos of cutting-edge AI models.

🤖 New Architecture: Server-Hosted LLM

The chat functionality has been re-architected to use a powerful, server-hosted Language Model (LLM) powered by a Python FastAPI backend. This replaces the previous implementation that used Transformers.js for client-side inference.

Key Changes:

• FastAPI Backend: A new backend service built with FastAPI hosts a Hugging Face text-generation model (e.g., mistralai/Mistral-7B-Instruct-v0.2).
• GPU-Powered: The backend is designed for deployment on a GPU server for high-performance inference, with Docker support included.
• Streaming API: The backend exposes a /chat endpoint with streaming support, allowing for real-time, token-by-token responses in the UI.
• Decoupled Frontend: The React frontend now communicates with the FastAPI backend, making it a lightweight and highly scalable client application.

🛠️ Development Setup

This project now consists of two main parts: the React frontend and the FastAPI backend.

1. Backend Setup (FastAPI + Hugging Face)

The backend runs a Hugging Face model on a Python server. It requires a machine with a GPU for optimal performance.

Prerequisites:

• Python 3.10+
• An NVIDIA GPU with CUDA drivers installed.

Instructions:

1. Navigate to the backend directory:

   cd backend-fastapi

2. Create and activate a virtual environment:

   python -m venv venv
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Configure the model: Create a .env file in the backend-fastapi directory and specify the Hugging Face model you want to use:

   MODEL_ID="mistralai/Mistral-7B-Instruct-v0.2"
   

5. Run the backend server:

   uvicorn main:app --host 0.0.0.0 --port 8000

   The server will download the model on the first run, which may take some time.

2. Frontend Setup (React)

1. Navigate to the project root directory:

   cd /path/to/MLAI-Stack

2. Install dependencies:

   npm install

3. Run the frontend development server:

   npm run dev

   The frontend will be available at http://localhost:5173 and will connect to the backend running on port 8000.

🐳 Docker Deployment (Backend)

The FastAPI backend can be easily deployed as a Docker container, which is the recommended approach for production.

Prerequisites:

• Docker
• NVIDIA Container Toolkit (for GPU support)

Instructions:

1. Navigate to the backend directory:

   cd backend-fastapi

2. Build the Docker image:

   docker build -t llm-backend .

3. Run the Docker container with GPU access:

   docker run --gpus all -p 8000:8000 -v ./huggingface:/app/huggingface llm-backend

   • --gpus all provides the container with access to all available GPUs.
   • -v ./huggingface:/app/huggingface mounts a local directory to the container to cache the Hugging Face models, preventing re-downloads on container restarts.

🔗 Live Demo

Visit the live application to explore the interactive ML demos and portfolio content.

📞 Contact

Carlos Gonzalez Rivera

• Email: cargonriv@pm.me
• LinkedIn: Connect with me
• Portfolio: Live demos and project showcase

---

This project demonstrates the intersection of machine learning research, software engineering, and user experience design. Each component is built with production-quality standards and serves as both a functional tool and a learning resource.