Luke Murray is an ISM Service Lead at HSO and a Microsoft MVP in Azure based in Hamilton, New Zealand. With over a decade of experience, he specializes in designing secure, scalable cloud solutions using Microsoft Azure, DevOps practices, Infrastructure as Code, and cloud-native architectures, while actively contributing to the global Microsoft community.
41 articles from this blog
Agent Package Manager (APM) manages AI agent dependencies like skills, prompts, and MCP servers, ensuring reproducibility and governance for Azure engineering teams.
A technical deep-dive into fixing Drasi's Azure AI Search backend crash-loop by tracing a root cause through Kubernetes and Let's Encrypt.
Drasi's MCP Reaction gives Copilot live data feeds by exposing continuous queries as subscribable MCP resources, pushing updates instantly.
Learn how to keep RAG pipelines fresh with Drasi's SyncVectorStore, which syncs embeddings continuously via change feed instead of nightly reindexing.
Explore using Drasi and Cypher for drift detection in AKS clusters, with insights on which rules work and which don't.
Drasi, a CNCF Sandbox project, enables continuous queries to detect absent events, replacing polling-based systems.
Building an AI-envisioning workshop tool on Azure that keeps evidence attributable and decisions accountable through a governed pipeline.
Discover how Azure Monitor SLI alerting can be deployed as infrastructure as code, despite initial challenges, by using metric alerts.
Azure Monitor SLIs on Managed Prometheus: undocumented gotchas, including the critical customdefault namespace, for building SLIs with infrastructure as code.
A guide to designing user-centred reliability with Azure Monitor SLIs and the Azure SRE Agent, using a payment service example.
This article discusses retiring a legacy dispatch layer in a codebase using the strangler fig pattern, migrating one domain at a time.
Building Execution Ledger, an open-source durable workflow orchestrator in Rust on Azure Container Apps, focusing on step-level recovery and idempotency.
Treating prompt changes like code deploys with eval gates to prevent silent failures in LLM-backed features.
Explains common failures when switching to reasoning-tier models in Azure OpenAI and how to avoid them.
Analysis of a system where AI agent failures silently triggered template fallbacks, hiding three critical bugs that made the system appear functional while broken.
Lessons learned building a production multi-agent service using Microsoft Foundry hosted agents, covering compute choices, protocols, and agent design.
Troubleshooting a version deadlock between AKS managed Gateway API and ALB controller, resolved by manual Helm installation.
Ringed deployment pipeline using Azure Developer CLI, Bicep, Helm, and ArgoRollouts for safe progressive delivery on AKS.
Architecture of an open-source World Cup 2026 fan intelligence platform using AI agents, Drasi, AKS, and Azure services.
A guide to deploying LiteLLM as a self-hosted LLM gateway on Azure Kubernetes Service using azd and Bicep with private networking and caching.
