Master’s Thesis Research
Policy-driven, Kubernetes-based resource management across multiple clouds, developed with Krateo PlatformOps and published as a journal paper.
Thesis
- “A Policy-Driven Kubernetes-based Architecture for Resource Management in Multi-Cloud Environments”, University of Trento, in collaboration with Krateo PlatformOps.
What I did
- Integrated Open Policy Agent as a Kubernetes mutating webhook to enforce latency-SLA, GDPR data-residency and carbon-aware scheduling policies, contributing to a system achieving up to 13% CO2 emission reductions vs. round-robin baseline scheduling.
- Developed Helm templates that dynamically select the best-fit VM size, gate provisioning on cloud provider and scheduling time, and drive VM and networking infrastructure creation across AWS, GCP and Azure via native Kubernetes Operators.
- Architected and deployed an MLOps pipeline serving ~15 PyTorch carbon-forecasting models using MLflow and KServe on Kubernetes.
Tech stack
Kubernetes · Open Policy Agent · Helm · AWS · GCP · Azure · MLflow · KServe · PyTorch