All architectures
Proposed DesignBanking AIOpenStack / hybridAWSMLOps and CI/CD

Continuous AI Improvement Workflow — OpenStack to AWS and Back

A proposed target design. Components shown are planned or conceptual unless stated otherwise.

A ten-step improvement loop that keeps live customer data on private OpenStack and moves only approved, de-identified datasets to AWS for training and staging.

Overview

  • Steps 1–3 on OpenStack: production usage and monitoring, selection of representative cases with human feedback and corrections, de-identification and data-quality validation.
  • Steps 4–6 on AWS: encrypted transfer of approved data to S3, reproduction of the current baseline, Llama / LoRA fine-tuning and evaluation in SageMaker, and staging validation on EKS/ECS.
  • Steps 7–8 in GitLab: automated tests, security and compliance checks, model evaluation review, human approval and release packaging into a private artifact store.
  • Steps 9–10 back on OpenStack: GitOps deployment with Argo CD, canary/regression tests, RAG re-indexing, promote or roll back, and monitoring that feeds the next cycle.

Business use case

Lets a regulated organisation improve its AI from real outcomes without moving live customer data out of its private cloud.

My role

Architecture design and documentation.