AutoML: Dataset Cleaner & Analyzer
End-to-end automated ML workflow - dataset upload → cleaning → EDA → training → predictions - with exportable PDF reports and a dataset-aware chatbot.
Building end-to-end ML pipelines from scratch is tedious - data scientists spend 80% of their time on preprocessing and EDA, not modeling. There was no low-code tool that could handle the full cycle while staying transparent about what it's doing.
Built a Flask web application that walks users through each stage sequentially. Each step is modular: the cleaning module handles nulls, outliers, and encoding; the EDA module generates statistical summaries and visualizations; the training module runs multiple algorithms and selects the best. The Gemini AI chatbot is given the dataset schema and sample rows as context.
- →Full upload-to-prediction pipeline in under 5 minutes for most datasets
- →Exportable PDF reports with EDA plots and model comparison tables
- →Downloadable pipeline artifacts (custom + baseline) for reuse without retraining
- →Dataset-aware chatbot that answers questions in natural language using actual column values
- 01Structuring a Flask app for sequential, stateful workflows requires careful session management
- 02Gemini API context windows need trimming - sending raw CSV data verbatim fails; sampling + schema works
- 03PDF generation with ReportLab is verbose but gives precise layout control for reports