Data and AI platform

MLOps and Data Platform Foundation on OCI

Created foundations for MLflow, OCI Object Storage, Data Flow/Spark, artifact tracking, and reproducible data/ML workflows.

Case-study indexHome

Role

Platform builder

Stack

MLflow / Data Flow

Outcome

AI-ready base

Problem

Data and ML workloads needed repeatable training, model tracking, private execution, and artifact flow.

Context

The platform needed AI/MLOps foundations that could support controlled execution, reproducible artifacts, and reliable data movement on OCI.

My ownership

Owned platform foundations across OCI Data Science, MLflow, Object Storage, Data Flow/Spark, image/artifact flow, logging, and delivery governance.

Architecture / delivery approach

Connected dataset storage, compute/runtime execution, model tracking, artifact handling, and operational checks into a repeatable foundation for ML and data workloads.

Outcome

Enabled AI-ready workflows with controlled execution, artifact flow, logging, and delivery governance.

key decisions

  • Separate datasets, runtime images, experiments, and model artifacts as first-class platform concerns.
  • Keep execution private and observable before scaling adoption.
  • Use repeatable delivery patterns instead of one-off notebook or job setup.

architecture examples

representative example

Experiment and artifact lifecycle

Lifecycle for reproducible data and ML work.

representative example

MLOps readiness checklist

Checks before treating a data or ML path as platform-ready.

  • Dataset location defined
  • Run metadata captured
  • Artifacts versioned
  • Operational signals reviewed

portable standard

OCI is the proven reference platform here; the MLOps operating model can be adapted by mapping storage, compute, experiment tracking, and CI/CD primitives in another provider.

OCI Data ScienceMLflowData FlowSparkMySQLTerraform