Strategy, architecture and hands-on delivery for AI systems that are useful, governable and built to survive production.
Explore↓The short version
I work at the point where executive ambition meets engineering reality. My background spans enterprise architecture, Azure, Power Platform, Copilot Studio, Microsoft 365 Copilot, Graph Connectors, agentic AI and technical program leadership.
At Microsoft, I help organisations translate emerging technology into secure, adoptable capabilities. Outside the slide deck, I build: agent tooling, MCP integrations, governance patterns, cloud runtimes and products that expose the real trade-offs.
Public portfolio
Only original public work is shown here—forked repositories are excluded. The projects go beyond chat into identity, actions, quality and observability.
A Dataverse-native observability solution that reconstructs Copilot Studio conversations, attributes users and exposes latency, reasoning, tool calls and correlated flow runs.
A governed, stateless ServiceNow catalog integration for Copilot Studio—designed around per-user identity, least privilege and operability.
Quality engineering for enterprise agents: architecture visualization, rule-based validation, solution checks, eval coverage and transcript analysis.
Practical Application Insights assets for operating Copilot Studio agents: reusable KQL queries, an Azure Workbook and review-first alert rules.
Selected private work
Private and organisational work is described only by its architectural essence.
A governed design for an agent-enabled company: deterministic orchestration, scoped workflows, human decision gates, audit state, risk tiers and accountable ownership—separated cleanly from its Azure runtime.
Operating model · Responsible autonomy · Observability · IaCAn evolving web and mobile product that turns household work into clear agreements. The build covers product discovery, multi-tenant architecture, Azure deployment, feedback loops, telemetry and continuous security hardening.
Product thinking · Full-stack delivery · Cloud operations · SecurityHow I lead
Start with a process owner, a baseline and a decision. AI is an option—not the objective.
Give agents the minimum authority needed, with identity, evidence, human oversight and a safe failure path.
Build reusable patterns for intake, delivery, evaluation, observability, cost and decommissioning.
Use AI to accelerate engineering while keeping architecture, review, tests and production ownership unmistakably human.
Where I create the most value