Use C5 instances for image processing, natural language processing, and more.
Machine learning in production
Use G4dn instances for training or serving machine learning models in production.
Large-scale inference
Use Inf1 instances to efficiently scale realtime and batch inference workloads.
Scalable
Cluster autoscaling
Let Cortex spin up more instances when load grows and spin them down when not in use.
Workload autoscaling
Customize the autoscaling behavior of each workload to ensure efficient use of cluster resources.
Flexible
Any workload
Define custom Python functions or containers and deploy them as realtime, async, and batch workloads.
Any pipeline
Chain workloads seamlessly to create custom data pipelines.
Observable
Structured logging
Automatically stream logs from every workload to your favorite log management tool.
Metrics aggregation
Monitor your workloads with pre-built Grafana dashboards and add your own custom dashboards.
Affordable
Spot instance management
Spend less on EC2 instances by letting Cortex manage spot instances.
Multi-instance type clusters
Configure resources per workload to run each workload on the right hardware.
Built for AWS
Identity & Access Management
Cortex integrates with IAM to simplify authentication and authorization worklfows.
Elastic Kubernetes Service
Cortex runs on top of EKS to scale workloads reliably and cost effectively.
Billing & Cost Management
Cortex supports resource tagging to enable detailed spend visibility.
Frequently asked questions
Is Cortex an alternative to AWS Lambda?
Yes, for some use cases. Cortex is a better choice for compute-intensive workloads like large-scale model serving or data processing while Lambda is ideal for running lightweight event handling code.
Is Cortex an alternative to AWS Fargate?
Yes. Cortex is a more flexible platform with fewer limitations on the types of resources and workloads it can scale.
Is Cortex an alternative to AWS SageMaker?
Yes, for large-scale model serving. SageMaker is more focused on end-to-end data science workflows.
Where can Cortex be deployed?
In a VPC on any AWS account. Cortex depends on AWS services like EKS, SQS, and S3 in order to enable reliable and cost-effective serverless computing at scale.
Discovering Cortex has been a lifesaver, it is servicing half a billion API calls each month for us. The ease with which we’ve been able to deploy Cortex has facilitated rapid development across our team, enabling us to meet the needs of our highly demanding customers.