Role Summary
Codvo.ai is helping in migrating one of our valuable client's model operations from Iguazio to Red Hat OpenShift AI in the second half of the year. Both platforms are Kubeflow-based, but the migration is still a significant program: the risk sits in the model serving runtime, observability, and parameter continuity, not in model development itself. You will plan and execute this migration alongside Baker Hughes engineers, backed by Codvo's Red Hat OpenShift AI architect relationships and internal ROSA clusters.
Responsibilities
Inventory the Iguazio estate: pipelines, MLRun projects, Nuclio functions, feature store usage, serving endpoints, and their consumers.
Map each Iguazio construct to its OpenShift AI equivalent (data science pipelines, KServe serving, model registry, workbenches) and flag the constructs with no direct equivalent early.
Migrate model serving with parameter continuity: same inputs produce the same outputs, verified with recorded traffic replay before cutover.
Rebuild observability on the target platform: serving metrics, drift signals, logging, and alerting at parity with the current setup.
Set up GitOps-based deployment (ArgoCD) and CI/CD for models on OpenShift AI.
Write cutover and rollback plans per model or pipeline group, and run the cutovers with Baker Hughes engineers.
Validate parity with side-by-side runs and sign off each migrated workload with its owner.
Document the target platform runbook so Baker Hughes can operate it without Codvo after handover.
Must-Have Skills and Experience
12 years in MLOps, platform, or DevOps engineering with ML workloads.
Strong Kubernetes; OpenShift experience preferred.
Hands-on experience with at least one Kubeflow-based platform (Iguazio/MLRun, OpenShift AI, or vanilla Kubeflow).
Model serving depth: KServe, ModelMesh, Seldon, or equivalent; traffic splitting and canary rollout patterns.
Python plus solid CI/CD and GitOps practice (ArgoCD, Tekton, or Jenkins).
Observability tooling for ML: Prometheus/Grafana plus at least one model monitoring approach (drift metrics, data quality checks).
Nice to Have
Direct Iguazio production experience (strongest differentiator for this role).
Red Hat certification; Codvo completed its Red Hat certification recently and the team is active in Red Hat's partner channels.
ROSA or other managed OpenShift experience.
GPU serving and vLLM.
Air-gap OpenShift operations.
Team Shape and Lead Variant
The migration program is staffed as one Lead Migration Architect plus two to three MLOps Migration Engineers. The lead role adds the following on top of the engineer scope:
Own the migration architecture and the 6-to-8-month plan, including sequencing, environment strategy, and the parity acceptance criteria.
Run the technical relationship with Baker Hughes platform architecture and with Red Hat OpenShift AI architects.
Make the call on constructs with no direct OpenShift AI equivalent: rebuild, replace, or retire.
Report program status and risk to Baker Hughes leadership; this workstream is time-critical for them.
Requires 10+ years including prior platform migration leadership.
First 90 Days
Weeks 1 to 3: complete the Iguazio inventory and the construct mapping; publish the migration backlog.
Weeks 4 to 8: stand up the OpenShift AI target environment and migrate the first low-risk pipeline end-to-end.
Weeks 9 to 13: run the first production cutover with rollback tested, and lock the repeatable migration playbook.