Red Flags in the Ledger - hf26 Challenge

TinyGuard AI is a lightweight, AI-powered anomaly detection tool designed to help users identify unusual expenses and suspicious transactions in their CSV data. Small businesses, freelancers, and independent professionals typically manage their expenses using spreadsheets, but manually reviewing every transaction can be time-consuming and may lead to unnoticed irregularities.

TinyGuard AI simplifies this process by allowing users to upload a CSV file and select numerical features for analysis. The application currently features two AI-powered analysis sections: Isolation Forest and TabPFN. Isolation Forest detects unusual patterns in numerical transaction data without requiring labeled training examples, while TabPFN demonstrates tabular classification by predicting Normal/Review categories using a synthetic, labeled training dataset.

The tool also offers visualizations, a results table, and downloadable analysis results. TinyGuard AI aims to make AI-based data analysis accessible to users who may not have the expertise to build their own machine learning pipeline. Potential users include freelancers, small businesses, independent consultants, and bookkeeping teams who would benefit from a tool that helps them review expense data more efficiently.

A future version of TinyGuard AI could include recurring CSV analysis, automated reports, and integrations with accounting workflows. The project structure consists of app.py (main application), requirements.txt (Python dependencies), sample_expenses.csv (sample data for testing), and README.md (this file). The application runs on Python 3.10 or higher and utilizes a stack of Python, Streamlit, Pandas, Scikit-learn, TabPFN by Prior Labs, and Matplotlib.

Isolation Forest identifies unusual observations, while TabPFN showcases classification using labeled synthetic training data.