MindAlpha Serving is the model serving (online inference) service for model trained by MindAlpha (https://github.com/mindalpha/MindAlpha). It provides low latency neural network model prediction and supports both CPU and GPU instance types on AWS. After subscription, the service can be installed via helm chart on your EKS cluster.
MindAlpha Serving provides horizontal auto scaling based on custom K8s metrics and automatically balances load on heterogeneous cluster nodes with different instance types.
MindAlpha Serving product charges by vCPUs per hour for CPU pods and vGPU per hour for GPU pods.
MindAlpha Serving provides opensourced Golang client with automatic service discovery and load balancing: https://github.com/mindalpha/mindalpha-serving-go-client
For big data analysis and MindAlpha model training, please check out our EnginePlus 2.0 product: https://aws.amazon.com/marketplace/pp/B08YMV3TV5