Enterprise MLflow
Security and scalability in an enterprise grade service with experiment and model sharing.
Fully managed service for data scientists
Experiments can be shared with chosen users and groups. Models can be shared with chosen users and groups.
Automatic scaling of small to large amount of users
No fixed Cloud Costs, pay per use. Hundreds of thousands of experiments and models.
Completely modern cloud architecture
Serverless. Not limited by traditional databases such as MySQL or PostgreSQL.
Encryption everywhere
Including on wire encryption and at rest encryption.
Robust Security
Experience a completely reimagined MLflow.
Authentication
All access methods are authenticated, including MLflow Web GUI, REST API, or CLI. Federate to any corporate directory.
Authorization
Experiment and model authorization allow permission to create, view and manage runs model versions.
Audit
Automatic and mandatory logging to MLflow Tracking with a complete audit trail of model version.
High Availability
Resilience to Data Center Failure. Service is guaranteed to be up even if an Availability Zone fails. Automatic – no administrator action required. No downtime, no loss of data.
Completely Serverless
Available in three Availability Zones by default. Auto Scaling and Load Balancing, hence resilient to DDoS
Static Files
Global Availability through CloudFront. Low Latency Loading of compressed assets from closest PoP
Disaster Recovery
Resilience to Region Failure. Administrator can bring up service in backup region in under 15 minutes. Manual Action required. No loss of data, about 30 minutes of downtime.
Completely Serverless
DynamoDB Global Table maintains replica of MLflow database
Persistent DynamoDB
Serverless Application can be brought up in backup region within minutes
Available Integrations
InfinStor logs code, parameters, and output of cell to MLflow, and captures cell reruns with modifications. Better than simply storing Python code in Git.
