Five years on enterprise .NET. A loop wired through Claude, Copilot & MCP.
Filed under Software Engineering · AI-augmented Development · Portfolio kept in HTML, by hand.
See selected work →
Three things on the bench right now. The rest of the bio is on the résumé.
Payroll & compliance services, monolith → versioned NuGet, SonarQube gates, EF perf wins on hot endpoints.
L3 · since 2021End-to-end agentic build: UI in Claude Design, implementation in Claude Code. Worked example of the loop.
Personal · liveDeepLearning.AI track — reflection, tool use, planning, multi-agent collaboration, MCP integration.
Andrew Ng · 2025+An agentic workflow used in production every day — ticket triage through merge-ready PR.
Read JIRA, repo, related PRs through MCP. Context is engineered, not retyped.
Agent drafts a plan, queries DB for the regression, identifies the hot path.
Patch + tests proposed inline. I review, refine prompts, reject what's wrong.
SpecFlow regression, coverage check, SonarQube gate. Nothing ships unverified.
PR opened with ticket links, change summary, perf delta. Merge gate green.
The skill isn't prompt-typing. It's context engineering, prompt design, and disciplined verification of generated output so it ships safely.
Daily · production · since 2024Kathmandu to Melbourne — fifteen years in. Education and work, one column.
Where .NET clicked. Data structures, OOP, databases, networks.
Multithreaded archiving & e-discovery in .NET. REST APIs in ASP.NET; profiled leaks, tuned T-SQL.
Software engineering, cloud, data. Pathway into the Australian industry.
Fatigue-management app for field crews — still in operational use. Express + React + Azure SQL.
APJ HR localisation, monolith → NuGet, XML import/export, SonarQube gates, 80%+ test coverage, EF/SQL perf.
Reliable LLM agents: reflection, tool use, planning, multi-agent collab, MCP integration.
Weighted by how much I actually live in each. Type size = depth.
One project. Worth more than five.
An interactive atlas of Victoria's crime stats. UI in Claude Design. Implementation in Claude Code. A real test of the loop from blank canvas to deployable artifact — the work was prompt design, context engineering, and verification.
If you need an engineer who ships the product and owns the AI loop around it — let's talk.