AI Research Engineer

Michael P BradshawMSc

I design and build AI-driven and full-stack applications — most recently completing an MSc in Artificial Intelligence. My research project evaluates the advancements ViT architectures can provide over the best current CNN's used widely in AI archaeological anomaly detection. Turning the experimental methodology into an End-To-End automated pipeline in Python. I am currently working on transforming the learning from this project into a commercial C#/.NET product. Take a look at what I've been working on.

View my projects Get in touch
Diagram of self-attention across a grid of image patches, with one patch attending to its neighbours and two distant patches

Fig. 01 — self-attention across patch embeddings

Featured work

Terra-AID

Terra-AID

MSc research project and desktop application investigating whether Vision Transformers improve unsupervised, reconstruction-based detection of buried archaeological features in Sentinel-2 satellite imagery.

PythonPyTorchPySimpleGUIVision TransformerMasked AutoencoderCNNResNet50Remote SensingUnsupervised Learning
MEyeTunes

MEyeTunes

An iOS app that turns a personally significant date — a birthday, a first dance, a favourite summer — into a UK #1 singles playlist, picked from a home-screen menu of playlist types.

iOSXcodeC++UI DesignSpotify APIApple MusicDeezerMobile Development

See all projects →