MARIJA VELLA
PROJECTS
Podspectrix Ltd.
Scottish Young EDGE winners 2022
Back in 2020, Alex, Sibo and myself started working towards developing a solution to prevent the incidence of diabetic foot ulcers. This led to the inception of Podspectrix, a MedTech startup using imaging technologies and AI to give everyone the diabetic foot care they deserve. So far we have raised £200,000 in funding and we are looking for further funding to progress the development of the device. Visit our website for more information. Our partners include:
Reliable Image Reconstruction Techniques: Enforcing Measurement Consistency in Deep Learning Methods (PhD Thesis)
My PhD project focused on designing new algorithms for reliable image processing. The current state-of-the-art methods rely on AI which although offer more visually pleasing images, they cannot ensure measurment consistency in their outputs. This means that information present in observed measurements is lost after being fed into a deep network. The aforementioned optimization algorithms do not suffer from this issue, which is why they continue to see limited use in applications where measurement consistency is essential.
Motivated by these observations, I designed algorithms that exploit the advantages of both optimization and AI networks for image reconstruction tasks. These were successfully applied to MRI reconstruction, resolution enhancement (super-resolution) and hyper-spectral image reconstruction. For more details please view the publications page.
Machine Learning Factor Analysis (MSc Thesis)
In my MSc thesis I created a ML-based system which is capable of identifying correlations between the inputs provided to Moody's Analytics economic scenario generator model and the financial variables it generates. Principal Component Analysis was performed to produce a simpler, more understandable view of the yield curve data. ML was implemented to accurately establish the structure of the complex dependencies between the parameters of the interest rate model and the PC coefficients and variances.
IoT Solution for Traffic Light Control (BEng Thesis)
In this project, I implemented a real-time IoT solution which adjusts the traffic light timings controlling an urban signalised junction. Vissim was used to simulate traffic conditions and collect data from simulated sensors. To obtain real-time optimal timing, the sensor data was uploaded and optimised on Microsoft Azure. Please view my thesis for more details!