PodcastLightweight Feature PlatformClaimTecton is a feature platform that decouples feature engineering code from models and helps manage feature pipelines.2:57
Feature stores, and second thoughts in 2024
16 sessions
PodcastLightweight Feature Platform
Building a Python-Centric Feature Platform to Power Production AI ApplicationsClaimA feature platform connects raw data to transformed features for model training and online inference while bridging offline and online environments.5:19
AI Innovations: The Power of Feature PlatformsPushed backNikhil Garg says a feature store and a feature platform are different: a feature store mainly provides storage and serving, while a feature platform also includes computation and other end-to-end capabilities.44:51
PodcastRecSys at SpotifyPushed backA vector database is not always a separate downstream destination for embeddings; in some online use cases it and the feature store serve overlapping roles.38:08
PodcastAccelerating Multimodal AIClaimA multimodal feature store would hold large media files together with structured information and derived features used for training models.14:15
PodcastBigQuery Feature StorePushed backMalt's BigQuery feature-store approach is not appropriate for every use case because live feature computation, very fresh data, or data too large for memory may require other systems.44:39
PodcastGlobal Feature Store: Optimizing Locally and Scaling Globally at Delivery HeroPushed backThe participants disagreed over the definitions of a feature and a feature store before aligning on shared definitions.20:34
PodcastUnpacking 3 Types of Feature StoresPushed backSimba Khadder argues that a vector store and a feature store solve substantially different problems despite both working with data and sometimes embeddings.12:17
Building a Data Infrastructure for AI/MLClaimA lakehouse can support ML feature engineering, training-data generation, model training, and production model serving from a shared data repository.11:06
Chronon: Airbnb's Open-Source Data PlatformClaimAirbnb increased its use of features from 3,000 to nearly 30,000 in about three years.1:53
Data Engineering: The Missing Piece of Your Data Science PuzzleClaimModels generally depend on batch features, near-real-time features, and real-time features.1:53
How Feature Stores WorkPushed backSimba Khadder disputes the idea that a feature store is merely a database or cache for storing features.14:12
Putting the AI back in Medallion Lake DesignClaimSimon Whiteley says machine learning workflows may draw data from any lake layer, perform preparation and feature engineering, use a feature store, run inference and produce result tables.5:17
The Evolution of Lyft's Feature StorePushed backThe speakers distinguish manually declared ownership and dependency metadata from automated observation of who actually reads feature data, arguing that manual metadata alone can become insufficient.12:00
Building an ML Platform from scratchPushed backBen says feature transformations should ideally be centralized in SQL mesh or a similar modeling system, while Eric favors monolithic pipelines early and sees decoupling as a later maturity step.1:26:27