OpenCloudHub
A master’s thesis project exploring modern cloud-native MLOps platform engineering.
About This Project
OpenCloudHub is a research-driven MLOps platform being developed as part of a master’s thesis in Computer Science. The project explores how to build modern, cloud-native infrastructure for machine learning workflows—bridging the gap between academic experimentation and production-ready systems.
What We’re Building
A comprehensive MLOps platform on Kubernetes that demonstrates the complete machine learning lifecycle. The platform showcases end-to-end system architecture, infrastructure-as-code, and modern DevOps practices for both traditional ML and generative AI models.
Key Technologies:
- Kubernetes - Container orchestration
- GitOps with ArgoCD - Deployment automation
- MLflow - Experiment tracking and model registry
- KServe - Model serving infrastructure
- Prometheus/Grafana - Monitoring and observability
- Keycloak - Authentication and authorization
- Gateway API & Istio - Traffic management and service mesh
- CloudnativePG & MinIO - Database and object storage
- Terraform/Terragrunt - Infrastructure as code
Research Focus
This thesis explores several key questions:
- How do modern cloud-native technologies integrate into cohesive platforms?
- What does comprehensive Kubernetes platform engineering look like in practice?
- How do MLOps workflows drive platform requirements and design decisions?
- How can we implement security and observability from day one?
The platform is under active development. Core infrastructure components are operational in a local Kind cluster, with MLflow tracking and basic model serving capabilities working. Documentation will be expanded as the platform implementation progresses. Follow the progress in our Project.