Coldstart- A library for automatic data curation + feature engineering
summary
Coldstart is a Python package that automates data curation and feature engineering by running concurrent templated SQL against dialects like BigQuery via SQLAlchemy, staging a leftmost table, and producing a merged table or pandas DataFrame ready for scikit pipelines. The demo runs five queries concurrently; CVS Health uses ~40 queries to produce thousands of features. It supports freezing query templates, intelligent downcasting and out‑of‑core outputs, and guards against temporal leakage so it actually returns usable training sets rather than just examples.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:06Cold start problem
- 3:18From scratch to feature stores
- 5:13How Coldstart solves it
- 7:42Key features of Coldstart
- 9:28Query structure and tags
- 12:00Leftmost table and dependencies
- 13:24Live demo setup
- 15:53Running feature extraction
- 18:22Modeling with features
- 20:51Productivity and collaboration
- 22:35Roadmap and extensibility
- 23:38Q&A
related talks
26mDask-SQL: Empowering Pythonistas for Scalable End-to-End Data Engineering
17mOpen Source for Data Quality
35mData Processing Pipelines for Small Big Data
01h 09mNotebook to Pipeline: Hands-On Data Engineering w Python
45mWorkshop 3: Data Engineering - A Roadmap... by Ramon Perez [Part 3]
38mData Prep Kit: A Comprehensive Cloud-Native Toolkit for Scalable Da...
11mBig Data Moonshots and Ground Control
30mFailing Fast with Data: Bringing CI to the Land of Data