I qualified with a 1st Class Honours, and Outstanding Achievement, in Mechanical Engineering, from a reputable University. I enjoy solving difficult challenges. Professionally I have...Read moreI qualified with a 1st Class Honours, and Outstanding Achievement, in Mechanical Engineering, from a reputable University. I enjoy solving difficult challenges. Professionally I have worked in numerous roles within the biotechnology manufacturing industry, and am well accustomed to solving complex challenges, to which I apply first principles.
Academic experience using MATLAB. Most notably, in the following two projects:
- Classify patients with Alzheimer's Disease based on EEG data. Time-series EEG data was split into fixed intervals in anticipation of different deep learning model inputs being prepared. The first of three inputs was time-series data mapped between 0 and 255 according to amplitude. The second, coherence function applied to time-series data converted into the frequency domain, to understand connectivity between electrode channels at different frequencies. Coherence matrix values were then assigned a value between 0 and 255 according to level of coherence. AlexNet, the deep learning architecture selected, required image data type to be input. The first two input types were converted to greyscale images, according to cell values; the final input dataset involved saving graphs as images. Datasets were split using inter-subject and intra-subject approaches, and with various different training:testing splits. Moreover, model hyperparameters were optimised to produce the highest classification accuracy. As alluded to, a pre-trained (foundational) AlexNet model was used, from the MATLAB Deep Learning Toolbox.
- Classify car brands from images of cars. I created a custom algorithm for seam carving, where the quantity of image material removed was dependent on the nearest edge which is of a length which exceeds a predetermined threshold. Idea was that the algorithm would calculate distance between the desired object, and the nearest side of the image, and remove any space between. This was done to remove any information not relating to the desired image (i.e the background), with the expectation that this would prevent the model from learning undesired features. The code produced iterates across “i” and “k” which are vertical and horizontal distances from origin, to find the lengths of different edges. Where only edges of 1’s beyond a certain threshold were considered. For example, the dataset generated for one test had a threshold edge length of 10% the image length. Meaning that distances between edges and sides of the image were only computed for edges which had a length of over 10% of the image’s length. This was done to prevent small artefacts from reducing the amount of seam carving completed.
Professional experience in utilising VBA for a large biotechnology company, in creating and validating excel userform and database (through use of VBA) which creates and prints required paperwork when users enter “product”, “batch” and “production order”. Different sub-assembly tolerances and processes exist across different products. This was implemented as an alternative to using a software product which was no longer supported. In addition to this VBA was used to automate production planning through use of a priority system, where slips to customer committed date were made traceable at each sub-process within production. Moreover, VBA was also used for simpler tasks such as extracting data from a Lotus Notes database and counting the number of overdue jobs for each machine. It was noticed, even these seemingly insignificant tasks, saved the business £££s per week.