automl-gs

Why we included this project

automl-gs is a rare AutoML project because it gives you the code instead of a locked-in model. You point it at a CSV file, name the column you want to predict, and it runs random hyperparameter searches across TensorFlow/Keras and XGBoost, then writes out a native Python pipeline with the data transformations and model construction in readable functions. Since the output is a plain script rather than a wrapped artifact, you can see exactly how your data was processed, tweak the architecture, and drop it into your own workflow without staying tied to the tool. It also saves you the tedious cleanup on real tabular data, handling messy bits like datetime and categorical encoding, plus column names with spaces or parentheses that would normally need manual fixes. That combination makes it a strong first pass for citizen data scientists and engineers without a deep statistics background, and the self-documenting generated code doubles as a way to learn how typical preprocessing and model choices actually work.