Feast
Feast is an open-source feature store under Linux Foundation governance, enabling ML teams to store and serve features for offline training and online inference. With 293+ contributors and 12M+ downloads, it serves enterprises including Robinhood, Capital One, and Walmart.
- Company typePrivate
- Founded2018
- HeadquartersSan Francisco, United States
- Headcount1–10
- GTM typeB2B
- OfferingSoftware
What Feast does
Feast is an Apache 2.0 open-source feature store governed under the Linux Foundation AI and Data Foundation, founded in 2018 as a collaboration between Gojek and Google Cloud to eliminate feature engineering duplication across ML teams. The platform provides a unified abstraction between ML models and underlying data infrastructure, consisting of an offline store for historical feature extraction during training and an online store for low-latency serving during inference. The core product surface includes a Python SDK, Python and Go feature servers (HTTP/REST, Arrow Flight, gRPC), a CLI, a Web UI, a feature registry, multiple compute engines (Local, Spark, Snowflake, Lambda, Flink), a stream processor, and a Kubernetes operator. Feast integrates natively with 15+ offline stores (Snowflake, BigQuery, Redshift, Spark, PostgreSQL, Trino, DuckDB, Azure Synapse, Dask, ClickHouse, MSSQL, Oracle) and 15+ online stores (Redis, DynamoDB, Bigtable, Cassandra, MySQL, SQLite, Dragonfly, SingleStore, Hazelcast, ScyllaDB, Milvus, Qdrant, Couchbase), plus orchestration, observability, and lineage tooling (Airflow, Prefect, Kafka, Prometheus, Grafana, OpenTelemetry, MLflow, OpenLineage, dbt, DataHub, Amundsen, Great Expectations).
The business model is pure community-driven open source with no commercial pricing, no paid tiers, and no direct revenue. Distribution is entirely self-serve via GitHub, PyPI (12M+ downloads), and documentation. The community consists of 293 contributors and a 5.5K-member Slack channel, with named maintainers from Red Hat, Affirm, J.P. Morgan, and Google. Named enterprise adopters span financial services (Robinhood, Capital One, Adyen), technology (NVIDIA, Discord, Cloudflare, Salesforce, Twitter/X, IBM, Red Hat), retail and e-commerce (Walmart, Shopify, HelloFresh), travel (Expedia), and events (SeatGeek). Tecton, the founding company behind Feast, offers a separate commercial managed feature store service; Feast itself remains a Linux Foundation project with no commercial operations, no disclosed headcount, and no disclosed funding.
Feast firmographics
Firmographics- Name
- Feast
- Legal name
- Feast
- Website
- https://feast.dev
- Company type
- Private
- Founded year
- 2018
- Operating status
- Operating
- Headcount range
- 1–10 employees
- Short description
- Feast is an open-source feature store under Linux Foundation governance, enabling ML teams to store and serve features for offline training and online inference. With 293+ contributors and 12M+ downloads, it serves enterprises including Robinhood, Capital One, and Walmart.
- Ownership category
- akta.pro rank
Where Feast is headquartered
LocationHeadquarters
- HQ city
- San Francisco
- HQ country
- United States
- HQ region
- North America
Markets served
Feast business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Personnel, Technology or R&D, Operations, Others
Revenue model
- Open Source Software: Feast is an open source project under the Linux Foundation AI and Data Foundation. No direct revenue is generated from the open source software itself. Commercial interest and funding comes through related commercial ventures (Tecton, the founding company, offers a managed Feast/feature store service).
Go-to-market motion1 record
Distribution channels4 records
Marketing channels5 records
Feast product offering
Product offeringCore offering
Feast is an open-source feature store that enables machine learning teams to define, manage, validate, and serve features for production AI/ML systems. The platform provides a unified interface between models and data, consisting of offline stores for historical feature extraction during model training and online stores for low-latency serving during real-time inference, with native integrations to 15+ data infrastructure systems.
Product overview
Feast is an open-source feature store that enables ML teams to define, manage, validate, and serve features for production AI/ML systems. The platform provides a unified interface between models and data, consisting of two foundational components: an offline store for historical feature extraction used in model training and an online store for low-latency serving in production systems. The core offering includes the Python SDK for defining features, the Python feature server for REST-based serving, CLI tools for management, and a Web UI for exploration. Supporting components include feature registries for metadata management, compute engines (Local, Spark, Snowflake, Lambda, Flink) for materialization, stream processors for real-time ingestion, and role-based access control via OIDC and Kubernetes RBAC. Observability is provided through Prometheus/Grafana integration, OpenTelemetry support, SOX audit logging, and native data quality monitoring (Feast 0.64). The platform also offers integrations with MLflow and OpenLineage for lineage tracking, supports vector databases (Milvus, Qdrant) for RAG workloads, and provides MCP (Model Context Protocol) support for AI agents.
Differentiator
Problem solved
Functional benefit
Products and services
- Feast Feature Store The central open-source platform that enables ML teams to define, manage, validate, and serve features for production AI/ML systems. Provides a unified interface between models and data with offline store for historical feature extraction during training and online store for low-latency serving during inference.
- Python Feature Server HTTP endpoint that serves features with JSON I/O, enabling any programming language that can make HTTP requests to read and write features from the online store. Supports performance optimizations including worker configuration, connection pooling, async reads, and pre-computed feature vectors.
- Go Feature Server Alpha feature server implemented in Go for high-performance feature serving requirements.
- Offline Feature Server Feature server using Arrow Flight protocol for efficient transfer of historical feature data for model training.
- Registry Server gRPC-based server for managing the feature registry, providing centralized access to feature definitions and metadata.
- Feast Web UI User interface for viewing and exploring feature definitions, monitoring feature quality, and managing the feature store. Includes dedicated Monitoring page with filters, summary tabs, feature drilldowns, histograms, and time-series charts.
- Feast Operator Kubernetes operator for deploying and managing Feast feature servers on Kubernetes with support for high availability, auto-scaling, and production-grade configurations.
Quantifiable outcome
- Sub-2ms p99 latency for single-row online feature requests with pre-computed feature vectors
- +3 more outcomes
Companies that use Feast
Customer profileNamed customers15 records
Segments4 records
Ideal customer profiles1 record
Feast technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration40 records
AI capability6 records
Feature9 records
Feast partnerships and signals
Strategic signalPartnerships
Eight partnerships are on record, tiered core.
- MongoDBcoreMongoDB announced an Atlas integration with Feast open-source store as part of their AI features. This integration is part of MongoDB's strategy to address AI reliability and trust issues in agentic AI workloads, combining Feast's feature store capabilities with MongoDB's vector search and Atlas platform.
- SnowflakecoreFeast provides native Snowflake integration as both an offline store (for historical feature extraction used in model training) and online store. Snowflake has collaborated with Feast on quickstarts and customer enablement.
- Google Cloud Platform (GCP)coreFeast was founded in collaboration with Google Cloud and provides native integration with BigQuery as offline store and Datastore as online store. GCP has published getting started guides for Feast on their platform.
- Amazon Web Services (AWS)coreFeast provides AWS integration with Redshift as offline store and DynamoDB as online store. AWS has published open source guides for running Feast on AWS managed services.
- RediscoreFeast provides Redis as the primary online store for low-latency feature serving. Redis has collaborated on blog posts and documentation for building feature stores with Redis.
- KubeflowcoreFeast is integrated with Kubeflow as an official add-on for feature store capabilities. Kubeflow provides official documentation for Feast integration in ML pipelines.
- AzurecoreFeast provides Azure integration with Azure Synapse as offline store and Azure Redis Cache as online store. Microsoft has published guides for bringing feature store to Azure.
- Red HatcoreRed Hat engineers (Jeremy Ary, Edson Tirelli) are official project maintainers. Red Hat is focused on enterprise readiness and OpenShift/Kubernetes integration. Active development contributions to the open source project.
Scale indicators4 records
Recent moves6 records
Expansion highlights6 records
Feast competitors and assessment
Company assessmentDirect peers
- Tecton: Tecton is the founding commercial sponsor of Feast and offers a managed feature store service built on Feast. It is the most directly comparable commercial peer because it sells essentially the same product to the same enterprise buyer, just packaged with commercial SLAs and support.
- Hopsworks: Hopsworks provides an open-source feature store with both community and enterprise editions, targeting the same ML platform engineers and data scientists as Feast. It is the closest direct OSS competitor with overlapping feature store, model serving, and project capabilities.
Broad incumbents
- AWS (SageMaker Feature Store): AWS bundles an online + offline feature store into SageMaker for its cloud customers. It is a broad incumbent competing on platform integration rather than feature parity, and represents the largest bundled alternative for AWS-centric ML teams.
- Google Cloud (Vertex AI Feature Store): Vertex AI Feature Store is Google's managed feature store, integrated with BigQuery and the broader Vertex AI platform. It competes with Feast primarily in GCP-native shops and on convenience of integration.
- Databricks (Databricks Feature Store): Databricks offers a feature store as part of its Lakehouse Platform, integrated with Unity Catalog and MLflow. It is a broad incumbent competing on the strength of the surrounding Databricks data ecosystem.
- Snowflake (Snowflake Feature Store): Snowflake launched a feature store leveraging its data cloud as the foundation. It is a broad incumbent competing with Feast on data gravity — Snowflake customers with heavy data cloud usage may consolidate ML feature management into Snowflake.
- Azure Machine Learning Feature Store: Azure ML includes a managed feature store integrated with Synapse, Azure Redis Cache, and the broader Microsoft ML ecosystem. It competes with Feast in Azure-centric enterprise deployments.
Emerging players
- Rasgo: Rasgo is an emerging feature engineering and feature store platform focused on automated feature generation and AI-driven transformations. It overlaps with Feast on feature definition and serving but emphasizes AI-assisted feature creation rather than pure infrastructure.
- Molecula: Molecula (formerly FeatureBase) offers a feature store / data platform focused on real-time, sub-millisecond feature serving. It overlaps with Feast on online feature retrieval and is a competitor primarily for latency-sensitive real-time ML use cases.
Others
- Iguazio (acquired by McKinsey QuantumBlack): Iguazio (now part of QuantumBlack / McKinsey) provided an end-to-end MLOps platform that included feature store capabilities. It is an adjacent/thematic peer that competed with Feast in the broader MLOps stack before its acquisition.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat5 records
Key risks6 records
Key highlights6 records
Customer concentration
Feast social profiles
Digital presenceFeast financial estimates
Financial estimateRevenue estimate
Valuation estimate
Feast leadership team
Management profileNumber of profiles
Profiles5 records
Feast funding detail
Funding detailFunding overview
Funding rounds
Investors
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Feast M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Feast
What does Feast do?
Feast is an open-source feature store that enables machine learning teams to define, manage, validate, and serve features for production AI/ML systems. The platform provides a unified interface between models and data, consisting of offline stores for historical feature extraction during model training and online stores for low-latency serving during real-time inference, with native integrations to 15+ data infrastructure systems.
Is Feast a public or private company?
Feast is a private company. It is classified as nonprofit foundation owned and is currently operating.
When was Feast founded?
Feast was founded in 2018. It employs 1 to 10 people.
Where is Feast based?
Feast is headquartered in San Francisco, United States, in the North America region.
How does Feast make money?
One revenue line is on record: open Source Software.
Who are Feast's main competitors?
Direct peers on record are Tecton and Hopsworks. Broad incumbents are AWS (SageMaker Feature Store), Google Cloud (Vertex AI Feature Store), Databricks (Databricks Feature Store), Snowflake (Snowflake Feature Store) and Azure Machine Learning Feature Store. Emerging players are Rasgo and Molecula. Iguazio (acquired by McKinsey QuantumBlack) is listed as an others.
Does Feast have an API?
Yes. Feast provides a Python SDK for programmatically defining features, entities, sources, and transformations, as well as reading and writing features to offline and online data stores. An optional feature server (Python-based) serves features via REST API over HTTP with JSON I/O, enabling any programming language that can make HTTP requests to interact with Feast. The feature server exposes endpoints for get-online-features, retrieve-online-documents, push, materialize, and materialize-incremental operations. Additionally, an Offline Feature Server uses Arrow Flight protocol and a Registry Server uses gRPC. Developer documentation is at docs.feast.dev.