AUTODATABOT
AUG 2025 — SEP 2025
An end-to-end automated data-analysis and machine-learning platform that accepts raw datasets, executes automated cleaning and quality analysis, trains machine-learning models, and generates comprehensive analysis summaries.
OVERVIEW
AutoDataBot automates the entire data science pipeline from raw CSV ingestion through to model training and report generation — removing manual preprocessing, model selection, and evaluation overhead.
THE IDEA
Data scientists spend the majority of project time on repetitive, low-value tasks: cleaning data, selecting models, and writing evaluation reports. AutoDataBot automates this entire pipeline to let practitioners focus on interpretation and high-level decision-making.
WORKFLOW
KEY FEATURES
- Dataset upload and ingestion
- Automated data cleaning and imputation pipeline
- Data quality evaluation and exploratory reporting
- Machine learning model training workflow
- Multi-framework AutoML engine
- Model evaluation and metric scoring
- Cleaned dataset export
- Trained model persistence and export
- Automated analysis summary generation
ENGINEERING APPROACH
AutoDataBot orchestrates multiple AutoML frameworks (AutoGluon, FLAML) with a unified evaluation interface. SHAP values are computed post-training for model explainability. FastAPI serves the pipeline as a REST interface, and SQLite tracks experiment runs. The core challenge was designing a flexible preprocessing pipeline that handles heterogeneous CSV schemas without manual intervention.