|
James Dorricott
I am a machine learning lead at Abingdon Health, where I work on computer vision models for interpreting rapid diagnostic tests.
Before studying for an MSc in Artificial Intelligence, I worked in equity research at Exane BNP Paribas, a global equities business. Prior to that, I worked in equity analytics at the sell-side spinout Libra Investment Services.
I completed my undergraduate degree in physics at the University of York.
Email  / 
LinkedIn  / 
GitHub  
|
|
|
Abingdon Health
I developed AppDx®, an ML-powered test reader recently out of stealth mode. I am the co-inventor of the product and operate as the senior ML engineer and engineering lead.
|
|
AppDx® AI Test Reader
James Dorricott, Neeraj Adsul
Abingdon Health, 2021 - Present
[product launch] [UK patent]
AppDx® uses real-time object detection and image classification algorithms to automate the visual reading of lateral flow rapid diagnostic tests. The deep learning software operates on-device and can be integrated into healthcare apps using the AppDx® SDK.
|
|
MSc at Imperial
I obtained an MSc in Artificial Intelligence from Imperial College London, completing coursework in symbolic AI, machine learning, deep learning, computer vision, natural language processing, optimisation, algorithmic fairness, and ethics.
I worked on machine learning safety for my thesis and also completed a group software engineering project.
|
|
Advanced Input Domain Splitting for Neural Network Verification
James Dorricott, Elena Botoeva
MSc AI Individual Project, 2020
[report] [slides]
Domain splitting is an effective technique for the verification of neural networks with high input-dimensionality. By using mixed-integer programming to find optimal splits, this thesis improves on bisection with a method that provably simplifies the verification problem.
|
|
Anatomical Landmark Detection and Demonstration Interface (ALADDIN)
James Dorricott, Alex Gaskell, Maleakhi Wijaya, Harry Coppock, Olle Nilsson, Faidon Mitzalis
MSc AI Software Engineering Project, 2020
[code and demo]
ALADDIN is a platform for the automatic detection of anatomical landmarks using deep reinforcement learning agents. With the ALADDIN software, researchers can also log human demonstration data and use the expert policies to accelerate learning.
|
|
Libra and Exane
At Libra Investment Services, I worked with Apollo®, an advanced times series asset pricing model. My contributions led to Libra winning Man Group's top award for investment alpha, ahead of a number of sell-side banks.
At Exane BNP Paribas, I worked with the deputy head of research as a research product manager and was a member of the investment review committee. I also designed my own earnings season research product.
|
|