Former Student: Rami Cassia

Rami is studying for a PhD in the Mathematics of Information at the Department of Applied Mathematics and Theoretical Physics in Cambridge.  He took the MPhil in Machine Learning and Machine Intelligence in 2020-2021. 

What was your background before starting the MPhil?

Before starting the MPhil I did a BEng in Aerospace Engineering at The University of Manchester. During my three years at Manchester I gained a solid understanding of engineering mathematics, as well as programming and analytical skills, and took particular interest in fluid dynamics related courses. In my final year I completed a dissertation on morphing wing design. During that time, I also began reading up on ML/AI and how it could be used in fluid dynamic modelling as well as active flow control. I therefore decided to do an MPhil in Machine Learning and Machine Intelligence at Cambridge to gain the skillset that would enable me to work on these interesting interdisciplinary applications. 

What did you get out of the course? 

The course gave me the opportunity to learn in-depth, and implement, a broad range of machine learning, deep learning, and reinforcement learning techniques. Moreover, a significant portion of the techniques taught rely on the robust Bayesian framework, which inevitably improved my understanding of probability and statistics. In practical assignments we were taught to use machine learning in domains such as computer vision, speech recognition, natural language processing, and interactive systems, all of which were mind- broadening and insightful. The practical assignments also really improved my programming skills. 

What were the highlights? 

I especially enjoyed working on my MPhil dissertation, which was on using physics-informed neural networks for reconstructing magnetic resonance velocimetry images, and it was a pleasure working under the supervision of Prof. Matthew Juniper and Ushnish Sengupta. Another highlight was the absorbing set of computer vision practicals on augmented reality, 3D shape estimation, and depth estimation, all of which were a joy to implement. 

What have you gone on to do after the course and what are your longer-term plans? 

I am doing a PhD in Mathematics of Information at the Department of Applied Mathematics and Theoretical Physics. I am being supervised by Prof. Richard Kerswell and we are exploring the use of encoder-decoder methods to recover and time-step turbulent flow fields. In the long term I plan to conduct research in the automotive or aerospace industry on the use of data-driven methods for fluid modelling, or perhaps even work in computer vision. 

What advice would you give to people applying for the course? 

Having some introductory knowledge of machine learning prior to applying is obviously important. More important however is to clearly show in your application what you would like to bring to the table - i.e., that you have thought about how you would like to use machine learning after course completion, or what avenues of machine learning you are interested in researching. Finally, be sure to also review some mathematical concepts including probability and linear algebra. 

What advice would you give to people taking the course? 

Aside from the academics associated with the MPhil, there are many researchers using machine learning in other divisions of the Engineering department. So if you are unsure about what you would like to explore in your MPhil dissertation, try to reach out to them as well! This will help you exchange more ideas before finalising your research topic. Should you also choose to continue to PhD studies at Cambridge, reaching out will also help you thoroughly consider your PhD research topic and keep your options open when applying. 

29th May 2022