At Revolutio Pty Ltd., I developed satellite-image segmentation models in PyTorch for a commercial engineering application. I exported the models to ONNX and wrote C# integration code for the client’s existing .NET software, where they remain deployed.
At Project Expedite Justice, I built a Python and JavaScript data platform that replaced manual Google Workspace workflows. I designed its GCP architecture and developed pipelines that brought millions of documents into a searchable system used by investigators.
At Transparency International, I developed red-flag indicators and clustering methods for EU beneficial ownership data, along with an interactive retrieval dashboard for investigators and policy experts. Earlier, at Hyundai MOBIS, I developed and optimised computer-vision models for driver-assistance applications through deployment.
My work also includes research in machine learning and pattern recognition at Merck KGaA, and a PhD in physics and computational neuroscience. I independently implemented a published self-organising recurrent neural network model in C and Python, and built an open-source MCP server for processing invoices and receipts.

