Skip to main content
Prisme.ai can be effectively deployed on AWS, leveraging managed services to ensure scalability, high availability, and operational simplicity. This guide provides step-by-step instructions and best practices to deploy Prisme.ai on AWS infrastructure.

​
Prerequisites

Before you begin deploying Prisme.ai on AWS, ensure you have the following:
  • AWS Account with sufficient permissions to create and manage cloud resources.
  • AWS CLI installed and configured on your local machine.
  • Basic familiarity with Kubernetes and AWS services (EKS, RDS, S3, Elasticache).

​
AWS Resources Setup

Deploy Prisme.ai using the following recommended AWS-managed resources:
  • Service: Amazon Elastic Kubernetes Service (EKS)
  • Recommended Configuration:
    • 3-5 nodes, each with at least 4 vCPUs and 16GB RAM
    • Autoscaling enabled for peak load management
  • Best Practices:
    • Use managed node groups for automated scaling and updates
    • Ensure that your EKS cluster spans at least two availability zones for high availability
  • Deployment: 
    eksctl create cluster --name=prisme-ai-cluster \
  --version=1.28 \
  --nodegroup-name=standard-workers \
  --nodes=3 \
  --nodes-min=3 \
  --nodes-max=5 \
  --node-type=t3.xlarge \
  --region=eu-west-1
  • Service: Amazon RDS for MongoDB-compatible (DocumentDB) or PostgreSQL or self-managed MongoDB via EC2/EKS
  • Recommended Configuration:
    • 3-node cluster for high availability
    • Instance type: db.m5.large minimum
  • Best Practices:
    • Enable automated backups and Multi-AZ deployments
    • Secure access via VPC security groups
  • Note: DocumentDB is AWS’s fully-managed MongoDB-compatible service. If your workload strictly requires MongoDB features unsupported by DocumentDB, deploy MongoDB manually or use AWS Marketplace AMIs.
  • Service: Amazon OpenSearch Service
  • Recommended Configuration:
    • 3 nodes, instance type: m6g.large.search
    • Dedicated master nodes for stability
  • Best Practices:
    • Enable encryption at rest and node-to-node encryption
    • Regularly snapshot your data to Amazon S3
  • Service: Amazon ElastiCache for Redis
  • Recommended Configuration:
    • 3-node Redis Cluster (cache.m6g.large recommended)
    • Multi-AZ for failover
  • Best Practices:
    • Regular maintenance windows for upgrades
    • Use Redis with JSON and Search modules if required by your deployment (via Redis Stack on EKS if needed)
  • Service: Amazon S3
  • Configuration:
    • Create buckets for model storage, user uploads, and public assets
    • Use AWS CloudFront CDN for public assets
  • Best Practices:
    • Enforce bucket versioning and lifecycle policies
    • Apply least-privilege bucket policies
  • Service: Amazon Elastic File System (EFS)
  • Recommended Configuration:
    • Use for shared file storage in Kubernetes workloads (e.g., PVCs)
    • Performance mode: General Purpose
  • Best Practices:
    • Regular backups via AWS Backup
    • Security groups limiting access to your Kubernetes cluster nodes only

​
Step-by-Step Deployment Guide

1

Set up EKS Cluster

Set up a Kubernetes cluster using the EKS CLI or AWS Management Console:
eksctl create cluster --name=prisme-ai --region=eu-west-1
2

Deploy Managed Databases

Provision databases and storage:
  • RDS (DocumentDB or self-managed MongoDB/PostgreSQL)
  • OpenSearch Cluster
  • Redis via ElastiCache
  • S3 buckets and EFS volumes
3

Configure DNS and Networking

Configure your Route 53 domains for the following:
  • API: api.yourdomain.com
  • Console: studio.yourdomain.com
  • Pages: *.pages.yourdomain.com
4

Deploy Prisme.ai via Helm

Use Helm to deploy Prisme.ai into your Kubernetes cluster:
helm repo add prismeai https://helm.prisme.ai/charts
helm install prisme-core prismeai/prismeai-core --namespace prisme -f values.yaml
Ensure your values.yaml is correctly updated with your AWS resources’ endpoints and credentials.
5

Configure Ingress and TLS

Use AWS Load Balancer Controller and ACM for SSL/TLS certificates to secure your deployment and route external traffic:
kubectl apply -f ingress.yaml
6

Test and Validate

Perform initial health checks:
kubectl get pods -n prisme
kubectl logs <pod-name> -n prisme

​
Security Best Practices

IAM Role Management

  • Use IAM roles for service accounts (IRSA) for secure, fine-grained Kubernetes-to-AWS permissions.
  • Regularly audit roles and permissions.

Network Isolation

  • Deploy within private subnets and secure via VPC and security groups.
  • Use NAT gateways for controlled internet access from private networks.

Secrets Management

  • Leverage AWS Secrets Manager or HashiCorp Vault for managing sensitive configuration.
  • Regular rotation of database passwords and API keys.

Monitoring and Auditing

  • Enable AWS CloudTrail, Amazon CloudWatch, and Prometheus/Grafana for comprehensive monitoring.
  • Set up alerts for unusual activities or resource consumption.

​
Next Steps

Kubernetes with Helm

Deploy Prisme.ai with Helm on Kubernetes

Products Configuration

Configure Prisme.ai AI products

Operations

Learn how to manage operations efficiently