Available for 3–6 month contracts
AI engineer building agents people can actually trust.
From train controllers overseeing GE’s routing AI to enterprise agents at Hivemind Labs, I build the systems and the oversight that decides when they can act alone.
Adobe
Hi, I’m Jake.
Every project I’ve worked on comes back to one question: when should a person trust the machine? In 2017 I led research and interface design for ARTC’s Hunter Valley network control program, giving train controllers oversight of GE’s routing AI. With Adobe’s XDI team I ran enablement for 30+ enterprise teams across APAC. Today I’m technical co-founder of Hivemind Labs, a research-first AI lab, building enterprise agents and the framework that decides how much they’re allowed to do.
I work deep in agent code and in front of stakeholders, which helps with the two hard parts: working out what to build, then getting people to rely on it.
- Based
- Melbourne, AU
- Works
- Onsite or remote
- Status
- Available now
Three ways I help teams ship AI they can stand behind.
Usually as a fractional lead on one problem, for three to six months.
Agentic development
Agents that work in production, not just in demos. Tools, memory, orchestration and evals, from the prompt down to the infrastructure.
Stack Python, TypeScript, AWS serverless · Claude, OpenAI, DeepSeek, GLM · Claude Agent SDK, OpenAI Agents SDK, MCP, LangGraph, or a custom engine when it’s leaner
- Production agents and the runtime under them
- Tool design, memory and multi-agent orchestration
- Evals and test suites that catch regressions before customers do
- Serverless delivery on AWS, in Python and TypeScript
Human–AI interfaces
Where people review, steer and approve what AI does. More than a decade of designing that handover, from a neural-net routing console to agent approval queues.
- Review, approval and exception-handling interfaces
- Operator research with the people who supervise the system
- Visual, voice and agentic interfaces
- Design systems that hold up as the product grows
AI safety & governance
Oversight that lets AI move fast without losing control. With Earned Autonomy, agents prove they’re safe through checks and evidence before they’re allowed to act alone.
- Risk, governance, oversight and compliance design
- Pre-send checks, tested like code
- Earned Autonomy: statistical evidence before an agent acts alone
- Audit trails people can actually read
- Working knowledge of ISO/IEC 42001 and ISO 27001
Selected work
More than a decade at the point where people meet machines.
Writing
Research notes on agent memory, checks and oversight.
Let’s build something people can trust.
Open to one 3–6 month contract. Melbourne or remote, Australia.