VICTORIA, AUSTRALIA
Russell Thomas.
I work in supply chain and IT for an Australian manufacturer in regional Victoria.
Most of what I do involves solving practical problems. Sometimes that means suppliers and production. Sometimes it means software, data or IT. I tend to get interested when something doesn't work as well as it should, then keep pulling at the thread until I understand it properly.
Outside work I spend a fair bit of time experimenting with AI, particularly recurrent models. That started as curiosity about memory and has turned into a much larger research project than I expected.
I'm also involved in my local community, enjoy flying, and generally like learning how things work.
What I do
Supply Chain & IT Manager
Barker Trailers · 2012–present
Woodend, Victoria
My main responsibility is keeping production supplied while getting the best practical value from our purchasing spend. That means suppliers, pricing, alternatives, planning, shortages and the hundred small problems that appear when you manufacture things for real.
Over the years my role has spread into IT and systems because I kept getting involved in problems that sat between the factory, the ERP and the people using them. I tend to get involved where information isn’t flowing properly or people are fighting the tools they’re supposed to be using.
Some of that work has included:
- Helping implement and improve processes connected to our ERP, so information can move between jobs without being entered again.
- Building internal manufacturing and business tools that bring production, purchasing and operational information together.
- Developing simpler digital workflows for goods receiving, production reporting and data capture, starting with the people doing the work.
I’ve learned that lasting change takes more than installing software. If the people using it hate it, I need to understand what I’ve missed. I try things, listen when my assumptions are wrong, change them and keep going. Regular requests for improvements are a good sign that something is becoming part of everyday work. A little humour helps too.
How I work
I start by trying to understand the problem. I talk to the people closest to it, build something small enough to test and change course when the evidence says I’m wrong.
I like complicated problems, but I don’t think the solution needs to look complicated. Whether it’s a supplier issue, factory software or an AI experiment, I want something simple that works and that the people using it can make sense of.
I’m persistent with problems. Failure usually makes me more curious about what I’ve misunderstood. Pulling something apart, testing an assumption and finding out why it failed is often where I learn the most.
Independent AI research
I started experimenting with recurrent AI models because I was frustrated by systems rebuilding context and losing track of what had happened before. That led me to look at how a model carries information forward, and eventually to RWKV.
My current project, RWKV7 State Machine, asks: can the internal state of a frozen language model be made to behave like a small programmable computer? “Frozen” means the base model’s weights stay unchanged; the setup uses small trained support components around it.
In controlled experiments, I’ve demonstrated arithmetic, memory, stored programs, branching, halting and repeated execution. I set pass conditions in advance, test on held-out cases and use negative controls to check what the results depend on. Failed attempts are part of the record too.
These are internally evaluated experiments, not a general-purpose system. They have not been independently replicated or peer reviewed. I publish summaries of the results, methodology, controls and limitations on GitHub; the underlying implementation and research materials remain private.
Community
I live in a small community in western Victoria and believe in contributing to the place you live.
I’m currently President of the Kaniva Lions Club and Vice President of the Nhill Aero Club. With Lions, I’d like to keep our work practical and close to home, particularly helping people whose difficulties might otherwise go unnoticed.
Our club also runs Driver Reviver, giving travellers a coffee, a biscuit and a rest break. It’s a simple way to do something useful for the people passing through.
Background
My path into IT started with a traineeship in 2003. Before Barker, I was the local IT repair guy, working mainly with small businesses: setting up networks, helping with accounting software, building computers and sorting out everyday IT problems.
- Certificate IV in Leadership and Management
- Certificate IV in IT — Network Management
Outside work
I hold a CASA recreational pilot licence. I love flying, but I’ve enjoyed learning how to fly almost as much as the flying itself.
I’ve been pulling computers apart since I was a kid and still have the same habit of disappearing down rabbit holes to work out how things work.
Currently
At Barker, I’m focused on making manufacturing information and workflows simpler and more connected.
In my research, I’m working on stronger ways to connect language to program execution and read back the results, while testing where the approach falls short.
I’d like to hear from researchers, funders and organisations working on practical AI, manufacturing systems or unusual computing architectures.
Always happy to have a conversation.
If we share an interest, or you see an opportunity to help or collaborate, get in touch ↗