Hopsworks Feature Store
hopsworks.ai
What it is
Hopsworks Feature Store manages, reuses, and serves machine-learning features. It pairs an offline store based on Apache Hudi tables on HopsFS with an online RonDB key-value store. The platform can generate training datasets using feature values from a historical point in time, helping prevent data leakage. Its Feature Registry supports metadata search, automatic analysis and statistics, version comparison, and access control. Hopsworks tracks feature distributions, detects drift, and can send alerts. Listed integrations include Airflow, Apache Flink, Databricks, Snowflake, Spark, AWS SageMaker, Vertex AI, and Weights & Biases. Deployment options include managed SaaS, Kubernetes on AWS, Azure, or GCP, and air-gapped data centres. The installer documentation lists AWS EKS, Google GKE, Azure AKS, and OVHCloud, requires Kubernetes 1.27 or higher, and recommends at least four to five nodes. Security options described include encryption at rest and in transit, project-based access controls, two-factor authentication, and SSO. The free tier includes one project and community support; SaaS is pay as you go, while Enterprise pricing is custom.
Spec sheet
- Product
- Hopsworks describes its Feature Store as a managed layer for managing, reusing, and serving machine-learning features.Read 29 Sep 2026 · hopsworks.ai
- Integrations
- Listed integrations include Airflow, Apache Flink, Databricks, Snowflake, Spark, AWS SageMaker, Vertex AI, and Weights & Biases.Read 29 Sep 2026 · hopsworks.ai
- Security
- The security page describes encryption at rest and in transit, project-based access controls, and authentication options including two-factor authentication and SSO.Read 29 Sep 2026 · hopsworks.ai
- Support
- The Free and SaaS pricing tiers list community support; Enterprise lists a dedicated support team.Read 29 Sep 2026 · hopsworks.ai
- Founded
- 2017Read 23 Sep 2026 · hopsworks.ai
- Company
- Hopsworks says it was founded in 2017 and lists offices in Palo Alto, London, and Stockholm.Read 29 Sep 2026 · hopsworks.ai
- Certifications
- Hopsworks states that it is certified under ISO 27001 and SOC 2 Type 2 and is GDPR compliant.Read 29 Sep 2026 · hopsworks.ai
- Deployment options
- The documentation describes managed SaaS and deployment on AWS, Azure, or GCP Kubernetes, as well as air-gapped data centres.Read 29 Sep 2026 · docs.hopsworks.ai
- Feature monitoring
- Hopsworks says it tracks feature distributions, detects drift, and can send alerts.Read 29 Sep 2026 · hopsworks.ai
- Feature registry
- The Feature Registry offers metadata search, automatic feature analysis and statistics, version comparison, and access control.Read 29 Sep 2026 · hopsworks.ai
- Online and offline storage
- The Feature Store pairs an offline store based on Apache Hudi tables on HopsFS with an online RonDB key-value store.Read 29 Sep 2026 · hopsworks.ai
- Point-in-time training data
- The platform can generate training datasets that reflect feature state at a historical point to prevent data leakage.Read 29 Sep 2026 · hopsworks.ai
- Self-hosting requirements
- The Kubernetes installer page lists AWS EKS, Google GKE, Azure AKS, and OVHCloud, Kubernetes 1.27 or higher, and recommends at least 4–5 nodes.Read 29 Sep 2026 · hopsworks.ai
- Who it is for
- Hopsworks presents the platform for data scientists, data engineers, developers, platform engineers, security engineers, and administrators building production AI systems.Read 29 Sep 2026 · docs.hopsworks.ai
Plans and prices
Strong points and limits
Strong points
- Includes online and offline feature stores.
- Supports point-in-time training datasets.
- Tracks feature drift and can send alerts.
- Offers managed SaaS and self-hosted deployment options.
Limits
- Free plan is limited to one project.
- Self-hosting requires Kubernetes 1.27 or higher.
- Enterprise pricing is custom.
Who it suits
Hopsworks suits data scientists, data engineers, developers, platform and security engineers, and administrators building production AI systems. Teams can choose managed SaaS or documented Kubernetes and air-gapped deployment options.
Hopsworks Feature Store in depth
Read the full write-up
Hopsworks Feature Store combines feature management with registry and monitoring capabilities for machine-learning systems. It is a strong fit for teams operationalizing AI across data and platform roles that need both online and offline feature workflows. Its breadth favors shared production systems over teams seeking only a lightweight feature store.
Overview
Hopsworks brings feature storage, governance, and deployment choices together. Teams can use its managed SaaS offering or deploy on cloud Kubernetes, with an air-gapped data-centre option for Enterprise. That flexibility suits organizations with differing infrastructure and security needs, though self-hosting calls for meaningful Kubernetes capacity and operational commitment.
Key features
The offline store uses Apache Hudi tables on HopsFS, while the online store uses RonDB. Keeping both in one platform supports historical training workflows alongside online feature access. Point-in-time training datasets reproduce feature values from a past moment, helping prevent data leakage when assembling model training data.
The Feature Registry offers metadata search, automatic feature analysis and statistics, version comparison, and access control. These tools help teams discover and govern shared features as definitions change. Monitoring tracks feature distributions, detects drift, and can send alerts, making it useful for teams that need visibility into changing production inputs.
Integrations include Airflow, Apache Flink, Databricks, Snowflake, Spark, AWS SageMaker, Vertex AI, and Weights & Biases. Security includes encryption at rest and in transit, project-based access controls, and authentication options such as two-factor authentication and SSO. Hopsworks states ISO 27001 and SOC 2 Type 2 certification and GDPR compliance.
Pricing
Hopsworks uses a freemium model. The Free plan costs nothing and includes one project, Feature Store, Model Registry, and community support, with no credit card required. Its single-project ceiling makes it suitable for evaluation or a contained initiative, but restrictive for teams coordinating several projects.
The SaaS plan has custom pricing, billed Pay as you go. It includes unlimited projects, Feature Store and Model Registry, Model Serving, usage-based pricing, community support, and a platform SLA. This is the more suitable managed option for teams that need to expand beyond one project and add serving; the usage-based model means costs depend on use.
Enterprise has custom pricing and includes all SaaS features, on-premise and air-gapped deployments, a dedicated support team, custom integrations, and a guaranteed SLA. It is aimed at organizations that need deployment control or dedicated support rather than a self-service plan.
Platforms
Hopsworks supports API, Linux, self-hosted, and web access, with both managed and self-hosted deployment and both online and offline stores. The documentation describes managed SaaS and deployment on AWS, Azure, or GCP Kubernetes, as well as air-gapped data centres. The Kubernetes installer lists AWS EKS, Google GKE, Azure AKS, and OVHCloud, requires Kubernetes 1.27 or higher, and recommends at least 4–5 nodes. That is a substantial starting footprint for smaller teams considering self-hosting.
Who it's for
Hopsworks is designed for data scientists, data engineers, developers, platform engineers, security engineers, and administrators building production AI systems. It is particularly compelling for teams that need feature reuse, historical training data, monitoring, and deployment flexibility in one platform. A single-project free tier is less suited to evaluating multiple workloads in parallel.
Pros and cons
Pros: Online and offline stores plus point-in-time training datasets support both serving and leakage-conscious training workflows.
Pros: Registry search, analysis, version comparison, access control, and drift monitoring address feature governance and production visibility.
Pros: Managed, cloud Kubernetes, and air-gapped deployment options give organizations room to match infrastructure and security requirements.
Cons: The Free plan is limited to one project, which constrains teams evaluating or using the platform across multiple initiatives.
Cons: Self-hosting requires Kubernetes 1.27 or higher and a recommended 4–5 nodes, raising the infrastructure bar.
Cons: SaaS and Enterprise have custom pricing, so neither offers a fixed price for straightforward budget comparison.
Alternatives
For teams choosing among Feature Store Software, Hopsworks stands out for its combination of online and offline storage, registry, monitoring, and deployment choices. Consider these alternatives when their pricing or platform model better fits your needs.
Amazon SageMaker Autopilot is a paid API and web option with a free trial and pay-as-you-go pricing; its listed on-demand pricing has no minimum fees or upfront commitments.
Canal is a free, open-source option for API, Linux, macOS, and self-hosted environments.
Feast is a free, open-source feature store for API, Linux, self-hosted, and web environments.
JFrog ML Feature Store is a paid option with a free trial and API, Linux, self-hosted, and web support; pricing is custom.
Metarank is a free option supporting API, Linux, macOS, self-hosted, and Windows.
OpenMLDB is a free, open-source machine-learning database with standalone and cluster versions.
Feathr is a free option for API, Linux, self-hosted, and web environments.
Chronon offers an Apache 2.0 open-source plan at free, with no usage limits stated; it supports API and self-hosted environments.
Verdict
Choose Hopsworks if your team needs a production-oriented feature platform that connects online serving, historical training data, governance, and monitoring, with managed and self-hosted deployment paths. Look elsewhere if you need a fixed SaaS price, a free tier for multiple projects, or a self-hosted option with a smaller infrastructure footprint.
Verdict
Hopsworks brings feature storage, registry, monitoring, and deployment options into one platform. Consider the one-project free tier and the requirements for self-hosting when comparing options.
Where Specifiction places it
- Feature Store Software#3 of 207.7