In a nutshell: A custom portal with embedded MLflow UI provides ML teams a persistent bookmarkable URL for experiment tracking. The architecture combines a React frontend, Flask reverse proxy with AWS SigV4 authentication, and Application Load Balancer for secure, centralized access management via SSO integration.
For growing ML teams, integrating Amazon SageMaker AI MLflow Apps into a proprietary portal becomes a scaling question. A solution with a React frontend and Flask reverse proxy enables centralized access management, SSO integration, and easier onboarding without AWS console access for data scientists.
As Machine Learning teams grow, managing access to MLflow applications becomes a challenge: presigned URLs don’t scale for teams with dozens of data scientists, and access to the AWS Management Console causes administrative overhead. Teams relying on SSO-integrated internal portals need MLflow experiment tracking alongside other enterprise-wide applications through a single, bookmarkable URL.
The solution consists of four components: an Application Load Balancer as the central entry point, a React frontend as a branded portal with embedded MLflow UI via iframe, a Flask reverse proxy for authentication and request signing, and an Amazon EC2-based backend. The Flask proxy transparently handles AWS Signature Version 4 (SigV4) authentication so users don’t need to manage AWS credentials.
The React frontend provides a customized portal experience with corporate branding. The Flask reverse proxy signs all incoming requests with temporary AWS credentials, whether for UI paths or REST API calls. Both CI/CD pipelines and automation scripts can programmatically access MLflow REST APIs through the same proxy endpoints.
The implementation is deployed via the AWS Cloud Development Kit (AWS CDK). For production environments, HTTPS with SSL/TLS certificates via AWS Certificate Manager (ACM) should be enabled. This allows organizations to reduce onboarding time, simplify access management, and provide data scientists with a consistent user experience across all internal tools.
Source: aws.amazon.com