Skip to content
← BACK TO PROJECTS
02 / 09
02AI / DATA SYSTEM

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

DATASET
CLEAN
ANALYZE
TRAIN
EVALUATE
OUTPUT

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.