Skip to content

README.md

Latest commit

 

History

History
126 lines (98 loc) · 4.65 KB

README.md

File metadata and controls

126 lines (98 loc) · 4.65 KB

🤖 AutoML Web App with Clustering, Classification & AI Q&A (RAG + LoRA)

This is a powerful and easy-to-use AI web application that lets users:

  • Upload a dataset
  • Choose between Classification or Clustering
  • Train models automatically
  • Download reports and predictions
  • Ask questions about the data using AI-powered Q&A (RAG + LoRA)

Built with custom machine learning models, optimized for deployment on AWS, and a clean UI.

🔍 Key Features

📊 1. Clustering

  • Automatically finds the best number of clusters (Elbow & Silhouette methods)
  • Supports K-Means and Agglomerative Clustering
  • PCA visualization for easy understanding
  • Creates and downloads a PDF report + CSV results

🧠 2. Classification

  • Users can choose from four custom-built models:
    • Naive Bayes
    • Decision Tree
    • Random Forest
    • Logistic Regression (with automatic hyperparameter tuning when user selects best model)
  • Or, let the system automatically select the best model based on ROC-AUC
  • After training:
    • Download the trained model
    • View and download log records
    • Download a full training report
    • Ask the AI about prediction results via the Q&A system (CSV does not include direct predictions)

💬 3. AI-Powered Q&A (RAG + LoRA)

  • Ask questions about your uploaded dataset and model results
  • Uses Retrieval-Augmented Generation (RAG) with TinyLlama
  • Fine-tuned on your data using LoRA (Low-Rank Adaptation)
  • Fast retrieval with ChromaDB

⚙️ Tech Stack

Area Tools & Technologies
Backend Flask, PyTorch, LangChain, Transformers
Frontend HTML, JavaScript
ML Models Custom Naive Bayes, Decision Tree, etc.
LLM TinyLlama + LoRA
Vector Store ChromaDB
Deployment Docker, AWS EC2, S3
CI/CD Crontab (checks Github for updates hourly)

📦 Architecture

Here's a simplified view of the system flow:

User → Web UI (Flask) → Model Selector → ML Training/Prediction
↓ ↓
AI Q&A (RAG) S3: Model, Logs, Results
↓
TinyLlama + LoRA

🗂️ Project Structure


project/
├── src/
│   ├── app.py                   # Main Flask application
│   ├── lora_train.py            # LoRA fine-tuning on TinyLlama
│   ├── rag_index.py             # Embedding & indexing for RAG
│   ├── rag_qa.py                # RAG-based QA interface
│   ├── utils/                   # Helper modules and shared functions
│   └── models/
│       ├── classification_main.py   # Full classification workflow
│       ├── classification_model.py  # All classification model implementations
│       ├── clustering_main.py       # Full clustering workflow
│       └── clustering_model.py      # All clustering model implementations
├── templates/
│   └── index.html               # Frontend UI (Form, Chat interface)
├── static/                      # CSS and JavaScript files
├── models/                      # Trained models (saved to S3)
├── logs/                        # Log files (viewable/downloadable)
├── requirements.txt             # Python dependencies
└── README.md

👤 Author

namdarine - No-Code AI Engineer
🚀 Live App: https://automlplatform.tech/
🧑‍💻 Portfolio: https://namdarine.github.io
✍️ Blog (Medium): https://medium.com/@namdarine
I'm currently building and sharing insights about no-code AI systems and automation.

Passionate about making AI more accessible, and empowering users to build AI without writing code.

🌟 Vision

I believe that AI should be created, understood, and used by everyone - not just engineers.

This project is part of my mission to break down the barrier between people and AI by providing a no-code, accessible platform.
It reflects my core belief:

"AI should not be something controlled by a few.
It should be a tool that anyone can create with, lead, and understand."

🧠 Philosophy: "AI belongs to everyone."
💡 Mission: "Empowering AI Without Code." / "Making AI More Accessible."

This belief drives the brand identity behind namdarine and my long-term goal to design a future where AI is truly a digital right, not a technical privilege.

📄 License

This project is licensed under the MIT License.
See the LICENSE file for details.