The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) Feature Stores for Accelerating AI Development - #432
Nov 30, 2020
In this discussion, Kevin Stumpf, co-founder and CTO of Tecton; Willem Pienaar, engineering lead at Gojek and Feast Project founder; and Maxime Beauchemin, founder of Preset and creator of Apache Airflow, dive deep into feature stores. They explore how feature stores can accelerate AI development, streamline data management, and address operational challenges. The conversation highlights the evolution of these stores, their importance in automating workflows, and the collaboration needed between data engineers and scientists to maximize efficiency.
AI Snips
Chapters
Transcript
Episode notes
Airflow's Origin
- Max Beauchemin created Apache Airflow at Airbnb, inspired by internal Facebook tools.
- It aimed to improve data pipeline management and batch processing for data engineering.
Feast at Gojek
- Willem Pienaar led engineering for Gojek's data science team, focusing on ML for various products.
- Facing scaling challenges, they developed Feast, their feature store, for better data management and productionization.
Michelangelo at Uber
- Kevin Stumpf led Uber's Michelangelo team, creating an end-to-end ML platform.
- This platform streamlined ML workflows, leading to a surge in ML adoption across Uber's products.