Could you tell us a little about yourself, and about your studies before starting the MSc?
I studied BSc Chemistry at Imperial College. The chemistry department taught us some Python and encouraged us to use it to analyse laboratory data, which I found super-interesting. As a result, I took a Python course during the summer break after my first year as an undergraduate, and this gave me the skillset to take more analytical and computational modules, and a computational chemistry thesis project during the rest of my degree. I also looked for other opportunities to grow my programming skills. I joined a hackathon for scientists and enrolled in a tech mentorship programme. Luckily, my mentor offered me an 8-week machine learning project at their start-up. The learning curve was steep, but I put a lot of effort into this project and earned a 3-month summer internship, which allowed me to dive deeper and gain some foundations before I started the MSc in AI.
What attracted you to the MSc in AI?
I love the fact that this course is designed for people without computing backgrounds but who are interested in the field. I used to think that only those who’d studied computing for their undergraduate degrees could work in AI-related areas, but this course shows people from other backgrounds can also do well in AI.
I also like it that the course just lasts a year, and is condensed. It allows students to learn most of the main machine-learning topics within the first two terms, followed by a software engineering group project and a 4-month research individual project. I realised later on how useful those skills and knowledge are when I was applying for jobs.
What did you enjoy the most?
This course is highly selective, so you will be surrounded by very talented minds. It is amazing to work with them and learn from each other. I spent a lot of late nights with my coursemates to tackle problems and discuss solutions. The fun and collaborative community on the MSc AI is definitely something I miss a lot after graduation.
Regarding the taught modules, the department has provided many options for you to choose from. I enjoyed Robot Learning, taught by Dr Ed Johns, the most. His lectures were very well-explained and the coursework was fun as well. (Potential spoiler alert: we were asked to build a simulated robot agent to go through mazes using different techniques taught!)
I also really enjoyed my individual project, which I talk about below.
What did you find more challenging?
I was a bit surprised by the number of lectures, tutorials and pieces of coursework when I first started the course. However, throughout the course, as a result of the pace, my time management skills improved and I feel a lot more comfortable dealing with several deadlines at the same time. I find it useful to dissect large deadlines into smaller tasks and create my own deadlines for those tasks on the calendar. It helped me to organise my work and life better and feel calm even when a lot of things are happening at the same time.
Could you tell us about some of your achievements on the MSc that make you proud?
It was a little beyond my expectations that I was able to thrive so well, given that I hadn’t studied quite as much maths as others on the degree who had taken mathematics for their undergraduate degree. There was less maths on my chemistry degree, and owing to this I was worried I wouldn’t be able to handle the intensity of the course. But from my own experience, confidence is the key. I wasted some time self-doubting my ability at the beginning of the course, but I soon realised that as long as I made the effort and organised my time well, it wasn’t as difficult as I imagined.
So if you are also like me, interested in this MSc AI degree but have some concerns about your background knowledge, I would strongly recommend you give it a go and apply! You will never know your potential until you challenge it!
What did you do in your spare time?
I enjoyed going to the gym. During my time on the MSc AI, I found it very stimulating to do an intense workout in the morning before I started my work. The adrenaline kept me excited and motivated to work throughout the day. At weekends, I sometimes spent time outdoors with friends, or played sports with people.
Could you tell us about your individual project?
Everyone's project topic and experience can be very different. This is my experience.
Because I’m interested in cyber security, I reached out to Dr Sergio Maffeis, who works on security and machine learning. I proposed a project in the field of machine learning applied to cyber security, even though I wasn’t certain of the specifics. But I was lucky that one of Dr Maffeis’s PhD students was applying reinforcement learning to build autonomous cyber defence agents, in collaboration with the Alan Turing Institute—and I am a big fan of reinforcement learning. So I happily joined Dr Maffeis's research team.
My project was associated with an international cyber security competition, called the CAGE Challenge. The competition organiser provided a simulated network system, the reward function for each action and two types of adversarial red agents. Participating teams are asked to defend against the red agents and protect the network system. Because there are more than a hundred actions to take and 11 thousand elements to observe, I decided to contribute to our RL agent by simplifying the action and observation space of the RL agent to improve its performance.
I worked closely with two researchers from the Alan Turing Institute, one researcher from the National Cyber Security Centre and the PhD student in Dr Maffeis's research team. We had meetings every week to share our results and discuss the challenges we were facing. It was a great experience for me to not only learn technical knowledge from them but also understand how working in a research-orientated job would feel like. I really enjoyed my time working with them, and learnt a lot as well. In addition, I’m happy that I contributed to building the best-performing agent, with the support of the team. This reinforcement learning agent eventually won 3rd place in the CAGE Challenge! We also published a paper about our model at the CAMLIS conference, with me as the second author. I am also very grateful that the Alan Turing Institute sponsored me to attend the conference in Arlington, US.
The Department of Computing has a lot of opportunities available to students. They also have a lot of partnerships and sponsorship from many excellent institutes and companies. Group projects and individual projects gave us a lot of freedom to explore what we are interested in!
What have you been doing since you graduated?
After graduating, I joined a computer vision start-up, PoseAI, to work on their 3D motion capture AI research. I love doing research. This job allows me to work on the research side and apply the technology to commercialised products. I am given enough trust to work independently on my research, but also enough support from the team when I am stuck. Unlike academic-focused research, I also need to consider practicality and user experience. For example, we need to consider the phone camera angles that users are likely to use when collecting training data; or we need to prioritise the solution to avoid predicted foot position sliding on the floor because users can be easily unsatisfied by this detail.
I appreciate the skillset and opportunities MSc AI has provided me. I couldn’t have imagined myself working on a job like this a year ago.
Do you have any advice for prospective students?
I wasn't sure if my background could give me enough mathematical knowledge. But I realise now that believing in yourself and trying your best is the key to making things happen.
The second piece of advice is time management. It was a fast-paced course, so having a well-organised timetable for yourself is important.
Last but not the least, taking breaks effectively. You can easily lose track of time when debugging a piece of code. While it is good that you are focusing on your work, remember to take a step back and have a mini break to refresh your mind. An easier or smarter solution might come to you after a little break.