Almost a year after graduation from St John’s, we catch up with Shreshth, who graduated from St John’s in 2020 with an MSci in Natural Sciences. He is currently studying for an MSc in Machine Learning at UCL, with learning taking place predominantly online.
Describe Machine Learning (ML) in a nutshell. How was the transition from Natural Sciences to ML?
I like to think of ML as an alternative to the classical software engineering paradigm. Normally we write code explicitly telling computers what we want them to do. But some problems are too hard to program (eg recognising a cat in an image) or are tasks that humans don’t know how to do well (eg finding anomalies in bank transactions). Instead of programming a computer with set instructions, ML trains models using experience (usually lots of data) to do a particular task, such as classifying news articles, playing chess or driving autonomous cars.
The undergraduate physics syllabus doesn’t feature ML, but it was a great entry point because it trains maths and programming skills. I was also fortunate to find an ML-related Part III project that served as an introduction to the field. It was interesting to see that some concepts have a direct overlap with physics, eg how entropy in thermodynamics is used in information theory. The MSc cohort is very diverse in STEM subjects, and it was exciting to see different people’s skillsets complimenting one another during group projects.
What was it like studying for the MSc remotely from your family home in Solihull?
Remote learning definitely has its challenges, particularly having to learn how to separate out work- and home-life effectively to stay productive. Lockdown started just after I ran the Cambridge half-marathon (my first), and I have tried to keep that habit up as a way to get out the house and away from the screen. I’ve actually signed up for a full marathon this October to keep me motivated. In terms of student experience, the biggest challenge was making connections with the rest of the cohort. Thankfully the transition online was pretty seamless, though: we had a supportive online postgrad community on Discord, and for practical sessions we used a platform called gather.town. More than anything though, I felt grateful to have a supportive and safe home to go back to, which is a huge privilege. Being at home allowed me to become closer to my sister and parents again after living away these past four years.
Tell us about your next move. Will it involve more academic study?
Yes, I’ve really enjoyed the MSc – so much so that I’m starting a DPhil in ML (AIMS CDT) at Oxford in October. I’m excited by the practical applications opened up by AI recently, but there is still a sizable gap between research and practice. I’m interested in working on these technical barriers to deploying intelligent systems with confidence in the real world. Until then, I am doing my thesis with a UCL spin-out, which is trying to solve some of these problems. I am also exploring my interest in the tech ecosystem from the investor’s viewpoint through working part-time at a Venture Capital firm, where I am helping to develop a data-driven approach to sourcing promising new ventures.
