Document and information processing
Extract, validate and route information from documents and workflows while preserving checks for uncertain or exceptional cases.
AI & automation / Sydney & Australia
Put AI to work on a real business problem.
I help businesses identify repeatable work that software and AI can make faster, more reliable or cheaper, then build the integration into the systems people already use. The starting point is the business process, not the model.
Where it makes sense
Automate the work, not the pitch deck.
A useful AI project has a defined input, a decision or task to support, a way to handle exceptions and a measurable result. It also needs to fit the data, software and people already in the business.
AI agents and connected tools
Connect models to business APIs and controlled workflows so they can assist with real tasks, with sensible permissions and human oversight.
Operational data and prediction
Use data and machine learning where prediction can improve a decision, then build that result into the software used by the team.
Practical feasibility reviews
Assess the process, data quality, integration cost, accuracy requirements and fallback before investing in a large build.
Work delivered
Automation with an operational result.
70% automated approval
At Ofload, a document AI model automatically approved 70% of proof-of-delivery documents, removing manual processing work equivalent to three full-time roles. The figure describes workload, not jobs lost.
Read the document AI case studyMachine learning in logistics
My work included freight cost prediction, carrier recommendations and load optimisation, with models connected to the operational platform.
Read the platform case studyHuman and AI engineering costs
I built CodePulse to connect software activity with human effort and AI token usage, making the real cost of engineering work easier to understand.
Read the CodePulse case studyHow I approach it
Start small. Prove the benefit.
First identify the manual steps and the exceptions. Then establish how success will be measured and whether the available data is good enough. I prefer a bounded integration with a human review path over an impressive demo that cannot run reliably.
I can help with the engineering as well as the decision: APIs, data flows, model integration, deployment and the checks needed to make the result usable. I'm based in Sydney and work with businesses across Australia.
Common questions
Before investing in AI.
Do we actually need an AI agent?
Not necessarily. A deterministic integration or ordinary software workflow is often a better answer. I start with the task and its constraints, then choose the simplest reliable approach.
How do we avoid an unreliable AI demo?
Define an acceptance test, measure errors and exception rates, keep a review path for uncertain cases and integrate only after proving the process is worthwhile.
Can you build the integration, not just recommend a tool?
Yes. My work spans software engineering, APIs, machine learning and operational automation. I'm comfortable getting into the code and connecting the result to existing systems.
Let's work together
What do you want
to move forward?
A product to build. A team to lead. A technology decision to get right. Tell me where you are and what you need.