Max Barker
AI Engineer
Hello ! I'm Max. I'm an AI Engineer in Perth, Australia.
I have been working in the AI space since 2023, primarily focusing on agentic systems and other cool things.
Outside of work, I like building things that I find interesting. I also have pet rats.
The best way to contact me is via email or linkedin, lets chat!
Projects
Publications
Explainable Detection of Microplastics Using Transformer Neural Networks
M. Barker, M. Willans, D-S. Pham, A. Krishna, M. Hackett · AJCAI 2022
Applied Transformer neural networks to detect microplastics (polyethylene & polypropylene) in marine samples using reflectance micro-FTIR spectral data. Achieved 98.7% accuracy on marine polymers, outperforming traditional ML baselines.
Sample FTIR Spectra
Model Comparison
Domain Adaptation for Microplastic Detection in Real-World Marine Samples
M. Barker, M. Willans, T. Singha, M. Hackett, D-S. Pham · MDPI Microplastics 2025
This work extends a prior Transformer-based microplastic classifier with unsupervised domain adaptation to address the challenge of distribution mismatch between datasets. The source domain comprises 252 PE and PP spectra collected from marine saltwater environments, while the target domain consists of 169 PE and PP spectra from commercially sourced standard reference polymers. Both were acquired using reflectance micro-FTIR spectroscopy, though their distributions differ due to variations in substrate, environmental degradation, and spectral noise. Three domain-adaptation methods — Deep CORAL, DSAN, and DANN — were integrated within the Transformer architecture and evaluated on both datasets, as well as a full 58,804-spectra .map file. Deep CORAL achieved classification accuracies of 99% on the source domain and 94% on the target domain, while DSAN achieved 95% on both domains, with neither method requiring access to target labels during training.
Training Convergence — source (solid) vs target (dashed)
Method Comparison
Pixel-level Predictions on Full Marine Map (58,804 spectra)
Ground Truth (PLS-DA)
Transformer Predictions
