MLOPS in the Non-RT RIC rAPP
Budget: $8 – $15 USD
I’m building an MLOps pipeline within a Non-Real-Time RAN Intelligent Controller (NON-RT RIC) environment to support intelligent network optimization. The goal is to automate the lifecycle of ML models—data ingestion, training, validation, deployment, and monitoring—using Kubeflow and KServe as core components.
What I Need
MLOps architecture design tailored to NON-RT RIC workflows.
Integration of Kubeflow Pipelines for model training, retraining, and CI/CD automation.
Configuration of KServe for scalable, production-grade model serving inside the RIC ecosystem.
Assistance connecting the pipeline to network-metric sources and RIC policy output logic.
Support in debugging model workflows, serving issues, containerization problems, or RIC integration gaps.
Clear documentation on architecture, APIs, deployment steps, and how to update or retrain models.
Engagement & Workflow
Hourly-based engagement (not fixed price).
Live online working/debugging sessions to troubleshoot issues and walk through implementations.
Flexible timeline—no hard deadline; we iterate until functional and stable.
What I Need
MLOps architecture design tailored to NON-RT RIC workflows.
Integration of Kubeflow Pipelines for model training, retraining, and CI/CD automation.
Configuration of KServe for scalable, production-grade model serving inside the RIC ecosystem.
Assistance connecting the pipeline to network-metric sources and RIC policy output logic.
Support in debugging model workflows, serving issues, containerization problems, or RIC integration gaps.
Clear documentation on architecture, APIs, deployment steps, and how to update or retrain models.
Engagement & Workflow
Hourly-based engagement (not fixed price).
Live online working/debugging sessions to troubleshoot issues and walk through implementations.
Flexible timeline—no hard deadline; we iterate until functional and stable.