Solutions Engineer by day, AI tinkerer by night. Based in Sydney.
I spend most of my time figuring out how to make AI actually useful, not the pitch deck version, the "does this hold up when real people use it" version.
I'm deep in the weeds of agentic AI, not just chatting with models, but building and contributing to systems that run autonomously and do real work.
Right now I'm spending time with:
- Dex - An open-source personal knowledge system built on Claude. I contribute to the project and use it daily as a testbed for agentic patterns.
- OpenClaw and other autonomous agent harnesses - exploring how different frameworks handle orchestration, tool use, and long-running tasks.
- Custom MCP servers and integrations - connecting AI agents to the tools people actually use (Slack, calendars, CRMs) so they can do more than generate text.
The through-line is making AI more than a fancy autocomplete. Agents that plan, execute, recover from failure, and actually get things done.
- Clarity of intent matters more than cleverness - Agents don't read between the lines. If your specification is vague, your output will be too.
- Fluent doesn't mean correct - AI fails confidently. The skill is catching what looks right but isn't, especially at the edges.
- Orchestration is a design problem - Multi-agent systems only work when you size tasks to fit the harness and define clean handoffs between agents.
- You have to know the failure modes - Context degradation, specification drift, silent failures. You learn to recognise these fast once you've been burned.
- Context is the real unlock - Getting the right information in front of the right agent at the right time is what separates a demo from a system.
- Not everything needs an agent - Sometimes a script is the right call. Understanding the economics of what to automate is half the job.
Solutions Engineer at Pendo. I run technical evaluations, build proof-of-concepts, and help enterprise teams figure out what's actually going to work for them. Less slideware, more working software.
