AI Code Review Platform

The problem

Teams want help spotting risky changes early, but an AI reviewer that blocks merges on a hunch, or keeps code evidence forever, creates new problems. The review had to be advisory, auditable and governed by each organisation’s own policies.

What I built

• Architected and developed an Azure-based advisory review platform that combines deterministic structural analysis with Azure OpenAI semantic review of GitHub pull requests, with an Azure DevOps integration path.
• An event-driven .NET 9/10 backend with Azure Functions, Service Bus, ASP.NET Core and EF Core on Azure SQL, wired to GitHub Apps, webhooks and Checks; I also contributed to the CLI and VS Code review workflows.
• A React/TypeScript dashboard for reviews, findings by severity and type, policy metrics, repository onboarding and organisation-level AI settings.
• Organisation access controls, audited management APIs, signed policy synchronisation, Windows-protected policy caching and bounded retention of finding evidence.
• Optional Microsoft Agent Framework orchestration, tenant-scoped Azure AI Search RAG with citations and configurable second-model validation — findings stay advisory and never block a merge.