Smart Synthetic Data Generator

• 

  Basic Ui of the website

Inspiration In many industries, training ML models is difficult due to limited or sensitive data. We wanted to build a simple GenAI tool that can generate synthetic tabular data to fill these gaps — all without needing cloud credits or paid services.

What it does The app takes CSV data, learns its patterns using the SDV GaussianCopula model, and generates realistic synthetic data. Users can view, compare, and download the synthetic data — with automatic stats, charts, and PII detection.

How we built it We used: Streamlit for the web UI SDV (Synthetic Data Vault) for synthetic data generation Matplotlib for distribution plots Python 3.13, with open-source libraries only All components run locally and work within free-tier limits.

Challenges we ran into Getting the SDV model to handle various CSV types and missing values Making the UI simple yet powerful enough for dual datasets Testing without AWS cloud resources due to credit limitations

Accomplishments that we're proud of Fully working GenAI tool using only free tools Clean and beginner-friendly interface PII detection feature to promote ethical AI use Support for two datasets and visual comparison — all local!

What we learned How generative models like GaussianCopula work Streamlit best practices for dynamic UI The importance of balancing simplicity with real-world utility Creating privacy-aware AI tools

What's next for Smart Synthetic Data Generator Add domain-specific presets (e.g., healthcare, finance) Add AI-powered explanations of dataset structure Support multi-table datasets and more advanced models

Built With

• python
• sdv
• streamlit