$whoami— Engineering Manager · ANZ · Melbourne
Building reliable systems for the agentic-AI era.
Engineering Manager at ANZ, Melbourne. I write about distributed systems, quality engineering, and putting agentic AI to work across the software delivery lifecycle — with the evaluation and human-in-the-loop guardrails that keep it safe in production.
10+ years
engineering
Engineering Manager
ANZ Digital Channels
Melbourne
Australia
Open-source builder
agent infrastructure
Engineer of the Year
ANZ · 2025
Selected Work
all_projects →Devstrap
Infrastructure as code for declaratively provisioning and reconciling a macOS developer workstation.
View project →Deep Agent Action
A security-conscious GitHub Action that runs coding agents inside CI runners to plan, review, and open pull requests.
View project →openapi-go-mcp
Generate Go MCP servers from OpenAPI 3.x and Swagger 2.0 specifications, turning API operations into agent tools.
Explore case study →
Start Here
- Evaluating AI agents like software →
A practical view of evaluation, tracing, and human judgement in agentic systems.
- From OpenAPI to MCP with Go →
How an existing API contract can become an agent tool surface.
- A deep dive into JSON-RPC →
The protocol foundations underneath many modern developer tools.
- RAG is not vector search →
Why retrieval design needs more than a similarity index.
Featured Writing
- genaiagentic-ai
An AI Agent in Your CI: Self-Hosted Coding Agents with GitHub Actions
SaaS coding agents want your source code. Self-hosting wants your weekends. What if the agent just ran inside GitHub Actions?
- genaiagentic-ai
Containing AI Agents at the OS Level: A Hands-On Look at Microsoft's MXC SDK
Microsoft's MXC SDK wraps model-generated agent code in policy-driven OS sandboxes — what it isolates, how to wire it up, and where the early preview falls short.
- genaievaluation
Evaluating an LLM Agent Like Real Software: Observability and Evals with Langfuse
A vibe-check isn't a test. How to trace, score, and gate an LLM agent with Langfuse — and the silent escalation regression evals catch that a demo never would.
Editorial Pillars
Agentic Engineering
Agents, MCP, evaluation, memory, and safe human-in-the-loop systems.
Distributed Systems
Cloud-native architectures, events, protocols, and reliability.
- distributed-systems
- kafka
Engineering Excellence
Quality engineering, observability, testing platforms, and developer experience.
- testing
- quality-engineering
Building
Go, open source, and architecture walkthroughs from hands-on work.
- golang
- tooling
Recent Posts
all_posts →- mcpprotocol
MCP 2026-07-28: What Changes in the New Specification
MCP 2026-07-28 introduces a stateless core, multi-round-trip requests, extensions, stronger authorization, and full JSON Schema support.
- genaiagentic-ai
Building Robust Intent Classifiers with Generative AI
LLM-based intent classifiers route zero-shot, with reasoning and no per-intent training data. How to build one robust enough for production multi-agent systems.
- genaiagentic-ai
The Unseen Engine of Modern AI Agents: A Deep Dive into JSON-RPC
Every AI agent that calls a tool is speaking JSON-RPC under the hood. A look at the decades-old protocol quietly powering MCP and the modern agent stack.
- genaiagentic-ai
Model Context Protocol: The "USB-C" for AI
Anthropic's Model Context Protocol gives LLMs a plug-and-play standard for tools and data — what it is, how it works, and why it ends the N×M integration problem.
- mathematicsprogramming
Essential Mathematics for Better Programming
The maths that actually earns its place in everyday programming — Boolean algebra, logic, and the concepts that make code simpler, starting with De Morgan's laws.
// Now
I lead enablement engineering while building and writing about the systems that help teams deliver reliable software in an AI-native world.