Crafting Superior ML Pipelines with SapientML
summary
SapientML generates runnable ML pipelines and source code from tabular data and a spec, producing notebooks that include EDA, preprocessing, sampling, and model training and evaluation within minutes. It builds multiple candidate pipelines (top three models+preprocessing), evaluates them, and selects the best, often returning CatBoost or LogisticRegression models and outputs runnable Python notebooks. Integration includes Hugging Face Spaces and a pip-installable Python API; project reports over 1,000 generated pipelines from public Kaggle datasets.
AI-written summary · about
chapters (7)
- 0:00SapientML project intro
- 1:05AI talent shortage and AutoML
- 2:35AutoML tool challenges
- 4:02SapientML tech overview
- 5:26Demo on Hugging Face Spaces
- 10:39Python API on Colab
- 13:21Conclusion and repo
related talks
45mAutomate your ML workflows with pipelines
18mAutomate the boring ML stuff with pipelines
34mImplement MLOps practices with Amazon SageMaker
27mScalable Data Pipelines - an ML Approach
29mHow to create fully automated ML workflows with Amazon SageMaker Pipelines
15mMetamodelization and Automatic Generation of ML Pipelines
31mWhy Your ML Pipeline Needs an Agent (and How to Build One)
36mBuild prod ML systems with feature /training /inference pipelines