I made many internship applications this year, but the one I waited eagerly to hear from was Peratech. I heard of Peratech in the first year of my degree when I learnt about the company's invention of quantum-tunnelling composites (QTCs). I learnt about the phenomena of quantum-tunnelling briefly in our Nanoscience module in Year 1, and I have been interested in Peratech ever since. So, when the opportunity opened up for me to apply in the middle of Year 3, I was very excited. I thought that the computational internship was the perfect playground to learn and absorb a coding repertoire, building from my limited experience in Machine Learning (ML) and scientific computing.
The project that was my focus throughout the internship involved the implementation of AI-based concepts in the design of sensors. Initially, I spent time learning the code and writing supporting documentation to help me truly understand what was going on "under the hood". Following this, I was tasked with amending the code itself to make some important improvements that would significantly improve its speed. In addition to this, I implemented a storage method that reduced the amount of space taken up by training data for the AI model. As I undertook this project, I researched many machine learning frameworks to determine the one ideally suited to deliver the required performance. Through our implementations we observed improvements in training accuracy and a significant increase in speed.
I expected to enjoy experimenting with the code and new techniques most. But at Peratech, especially under guidance from Dr Tim Wiles, I experienced a lot of encouragement and a very constructive work atmosphere. I believe that is rare to find in R&D industries these days. That was the main highlight for me. The other highlight was seeing my improved framework reach the goal level of accuracy. Machine learning is a task that demands patience and focus, and it was great to see my work pay off.
Completing this internship has given me valuable insight into problem-solving. For example, after experimenting with a bottom-up approach to improve my code and training model's size and accuracy, I realised I needed to shift gears. I needed to think of approaching the problem in a different way, a more top-down approach. I believe that this issue I encountered during my project will be a valuable learning experience for me. But the icing on the cake for my internship was Dr Wiles. Dr Wiles offered me lots of advice and constructive criticisms throughout my internship, which I took on board. I found this very encouraging, and I intend to practice these as I take on more computational projects in my 4th year and beyond.
