Queensland Government · NRMMRRD · Jul 2025 to Sep 2026
AI-assisted workflow platform
A platform for assessing government applications. It processes the submitted documents, runs each case through a configurable workflow, and gives assessors LLM decision support they can override, with every step recorded for audit.
My role
Originated the project as a capstone and led the six-person team that delivered the proof of concept. Retained by the department as one of two engineers, then delivered the MVP: workflow engine, responsible AI features, stack selection, environments and CI/CD, within the allocated budget.
Stack
Next.js · NestJS · Azure OpenAI · RabbitMQ · Redis · MongoDB · Docker
What the platform does
- 01
Document processing
Handles the documents submitted with each application.
- 02
Workflow orchestration
A configurable engine that moves each case through its steps.
- 03
LLM decision support
Model output that informs the assessor's decision.
- 04
Human-in-the-loop override
Assessors review and can override every recommendation.
- 05
Audit trail and explainability
Decisions are recorded and explainable for public accountability.
From capstone to delivered MVP
Jul 2025
Capstone begins. Turned a machine-learning brief for application prioritisation into a broader workflow solution, leading a six-person Agile team.
Jun 2026
The proof of concept becomes a funded project. The department retains two engineers to build it.
Sep 2026
MVP delivered within budget. The department determines the platform's next steps.