I was engaged through Novus Strategy from January 2025 as Technical Architect on a UK home-buying programme, leading platform-level design decisions and driving architectural standards across the organisation.

My primary remit was the orchestration capabilities at the centre of the home-buying journey: the digital ecosystem connecting the parties to a residential transaction. The platform aligns with the PDTF (Property Data Trust Framework) schema, a UK standard for property transaction data interoperability. I flagged a structural risk with tightly coupling internal storage to an externally-controlled schema subject to frequent breaking changes, and shaped the integration approach accordingly.

The most significant architectural contribution was a unified authorisation service to replace logic that had drifted across eight services, each implementing access control differently and each holding direct read/write access to organisation and transaction data. The replacement supports three models in a single service: role-based access control, fine-grained permissioning, and relational authorisation that resolves a user's access to a transaction through their position in the organisational hierarchy and that organisation's role on the transaction. It builds on patterns I'd previously developed at Spa Space.

The greenfield design landed in two to three weeks. The work to keep the legacy estate functional alongside it took three months, an instructive cost-of-coupling that informed how I framed subsequent migration proposals to leadership. Under production-scale load the new service held to 300 millisecond response times against a baseline of 10 seconds in the legacy implementation.

I also designed and built an automated documentation and intelligence layer over the client's 390+ repositories: an MCP server with RAG vectorisation for AI-powered codebase interrogation, paired with execution-path analysis that generates living C4 and sequence diagrams. Engineering teams can now query the platform in natural language or generate up-to-date architectural documentation on demand. The capability has become a foundation for the organisation's AI-augmented engineering posture, which I'm continuing to refine.

More recently I've defined Minimal Viable Documentation (MVD) as a standard, with an AI-driven process that produces and maintains MVD against any repository in the estate. I've also researched and designed AI-driven performance and resource-consumption analysis for deployed services, integrating Azure DevOps, DataDog, and the vectorised codebase as a single evidence base for architectural decisions. On the strategy side I've recommended consolidating three overlapping services into one for maintainability, advised on deployment strategy, and supported critical development phases directly using Claude-augmented AI development.

Beyond delivery, I contributed to architectural governance, mentoring, and engineering uplift, advocating for proper refinement and standards adoption before sprint commitment, and helping shape how the organisation approaches AI-augmented development at team and platform level.