Dr Nick Wright

• Developed OmniCloudMask, a sensor-agnostic deep learning cloud and shadow masking tool, the most downloaded of these packages.
• Scaled the Sentinel-2 analysis pipelines to production, processing large-scale raster datasets across Western Australia's agricultural region.
• Developed OmniWaterMask, a hybrid deep learning and NDWI-based water body detector integrating spectral indices and vector datasets for robust farm dam monitoring.
• Developed BuildingRegulariser, a library for cleaning and regularising building footprints by aligning edges to principal directions and simplifying polygons.
• Developed Smoothify, which smooths raster output from segmentation and pixel-based classification so the derived polygons follow natural shapes such as water body outlines.
• Designed high-throughput deep learning workflows achieving ~45x inference speed improvements via asynchronous data handling, optimised tiling strategies and model architecture selection.
• Investigated and resolved false positive, cloud, haze, and artefact issues across multiple sensor types including Sentinel-2, Landsat, and drone imagery.
• Shipped the Sentinel-2 Grid Explorer, a public web map for locating and naming Sentinel-2 grid tiles.
• Delivered geospatial ML training for partner organisations including Landgate.
• Contribute to the DPIRD DMA team blog on Python, remote sensing and machine learning.

• Continues research collaboration following completion of the PhD.

• Moved into remote sensing and machine learning from 2019, applying deep learning to high-resolution Maxar Vivid 2.0 imagery with early fastai and PyTorch, and building the first Sentinel-2 analysis pipelines.
• Produced two open government datasets from that work, published on the WA Data Catalogue and used operationally for land and water management: Farm Dams of WA and Buildings of WA.
• Ran DEM modelling and surface flow analysis using Whitebox GAT.
• Before 2019, worked on drone survey and photogrammetry, hydrogeology, and GIS analysis.

• Embedded with Landgate's data team to develop machine learning workflows, create geospatial ML products, and deliver a course on applied geospatial machine learning.

Thesis: Deep learning for remote sensing: scalable, accurate, and sensor-agnostic approaches for environmental monitoring. Awarded a place on the Dean's List for outstanding doctoral theses. Three core contributions: high-throughput architectures and training dataset unification; sensor-agnostic training for cross-platform generalisation; novel prediction merging combining deep learning with spectral indices and vector datasets.

Dissertation: Hydrogeology and hydrochemistry of the unconfined aquifer of the Broome Peninsula. Completed as a separate honours year.