Serving Data for Production AI
Feast is an open source feature store that delivers structured data to AI and LLM applications at high scale during training and inference
[ADOPTERS AND CONTRIBUTORS]
[USE CASES]
SOLVE REAL PROBLEMS
Real-Time Recommendations
Serve personalized product and content recommendations with real-time user interaction features
Fraud Detection
Detect fraudulent transactions using historical patterns and real-time behavioral features
Risk Scoring
Calculate risk scores for financial services using consistent features across training and inference
Customer Segmentation
Create dynamic customer segments using consistent feature definitions across teams
[INTEGRATIONS]
CONNECT WITH YOUR STACK
OFFLINE STORES
ONLINE STORES
[GET STARTED]
START SERVING IN SECONDS
from feast import FeatureStore
# Initialize the feature store
store = FeatureStore(repo_path="feature_repo")
# Get features for training
training_df = store.get_historical_features(
entity_df=training_entities,
features=[
"customer_stats:daily_transactions",
"customer_stats:lifetime_value",
"product_features:price"
]
).to_df()
# Get online features for inference
features = store.get_online_features(
features=[
"customer_stats:daily_transactions",
"customer_stats:lifetime_value",
"product_features:price"
],
entity_rows=[{"customer_id": "C123", "product_id": "P456"}]
).to_dict()
# Retrieve your documents using vector similarity search for RAG
features = store.retrieve_online_documents(
features=[
"corpus:document_id",
"corpus:chunk_id",
"corpus:chunk_text",
"corpus:chunk_embedding",
],
query="What is the biggest city in the USA?"
).to_dict()[BLOG POSTS]
THE LATEST FROM FEAST
Extending Feast Observability: Offline Store Metrics and SOX Audit Logging
Feast now captures RED metrics for offline store retrievals and emits structured SOX audit logs for both online and offline feature access — closing the observability gap between serving and training paths.
Tuning the Feast Feature Server for Sub-2ms Online Serving
A practical guide to achieving low-latency, high-throughput feature serving with Feast on Kubernetes — from default configuration to production-grade performance with pre-computed feature vectors and benchmarks at every step.
Native MLflow Integration for Feast: Automatic Feature Lineage for Every Experiment
Feast now ships native MLflow integration : enable it in feature_store.yaml and every feature retrieval is automatically linked to the MLflow run that consumed it. No glue code, no manual tagging, full model-to-feature traceability.
[GET STARTED]
START BUILDING TODAY
Join our Slack
Become part of our developer community & get support from the Feast developers
Join Community