Twenty-five containerised AI applications running on a shared Azure Kubernetes Service cluster with Knative scale-to-zero, a private PyPI registry, and a file-ledger service giving sub-10ms blob discovery across the platform. Built and operated as the platform engineer responsible for everything from local Docker Compose developer environments through to production observability.
The work spans IFRS 17 financial extraction pipelines, retrieval over accounting standards and audit working papers, voice processing and unstructured document intake, anomaly detection on financial transactions, and a knowledge-graph compile architecture that treats wikis as the source of truth and RAG as the retrieval layer.