My path
I first became interested in machine learning in 2017, just after finishing my A-levels, through two formative experiences. The first grew out of my friendship with Wesley, who was then a computational biology PhD candidate at NUS. I had met him two years earlier, shortly after completing my O-levels, while trying to build a bioprinter at a local makerspace in Singapore. When he later saw an Instagram post about my attempts to study machine learning through Udacity, he reached out and worked with me on my first Kaggle competition: the 2017 Data Science Bowl, which involved detecting signs of lung cancer in CT scans. That project gave me my first exposure to Unix, training convolutional neural networks on GPUs, and classical methods such as naive Bayes and decision trees.
The second experience was attending CS6101, a course run by Professor Min-Yen Kan at NUS and based on Stanford’s CS231n. It gave me my first serious exposure to computer science and pushed me to think more deeply about the mathematics underlying machine learning. More importantly, it was where I met Eldric and Arun, who would both become lifelong friends. Before I had even started university, Eldric taught me propositional logic, linear algebra, and many of the other fundamentals on which my later studies would be built.
These experiences culminated in two projects that shaped how I learned. For the first, I implemented the generative adversarial network from Ian Goodfellow’s original paper on MNIST from scratch. That taught me how to read a research paper, work through its ideas, and turn them into functioning code. For the second, my CS6101 project, I used an LSTM to perform base-calling for nanopore sequencing and presented the work at NUS. This was around the time transformers had only just appeared, so there were no transformers involved—just recurrent networks, a great deal of experimentation, and a lot of learning.
All of this happened while I was completing my two years of National Service. To be honest, they were some of my most intellectually formative years. I had the freedom to read widely and pursue questions far beyond the demands of the A-level curriculum, and it was during this period that much of my broader thinking began to take shape. I read across philosophy, business, science, politics, history, and literature. I was also fortunate to work at headquarters, where I learned from majors, colonels, and generals about Singapore’s military doctrine. Those conversations developed my political awareness and capacity for strategic thinking in ways that a university course on the same subjects might not have done. I remain deeply grateful for the experience. It rekindled my interest in the humanities—particularly history and philosophy—and established a foundation for how I think today about technology, politics, and a changing world order.
My next major milestone came at the end of my first year at university, when I worked with Fatir Ansari and Harold Soh on improving the samples produced by generative models. That work eventually became my first scientific paper. The project also forced me to engage much more deeply with calculus. I began teaching myself real and numerical analysis and, in the process, discovered some of the beauty of mathematics—particularly ideas such as the generalized Stokes theorem. This became a major motivation for switching my degree from computational biology to applied mathematics. It also shaped my taste in mathematical questions: I became fascinated by problems such as Hilbert’s sixth problem and, more broadly, by the foundational relationship between mathematics and the physical sciences.
Then COVID arrived, and it is difficult to overstate how hard those years were. Lockdown, illness, personal tragedy, and upheaval within my family placed every part of life under strain. Some family members died; others had their lives irrevocably changed. The effects remained with me for years and eventually contributed to my decision to take a gap year from my PhD in 2025.
Even so, I completed my undergraduate degree at the top of my applied mathematics cohort. One of the most important experiences of that period was taking part in Google Summer of Code, where I worked with Kevin Murphy and Mahmoud Soliman on Kevin’s textbook. That opportunity led me to work with Emtiyaz Khan as an undergraduate intern at RIKEN and, eventually, to receive the Lijien Industrial Medal for the best thesis in applied mathematics.
Still healing from the difficulties at home, and determined to encounter more of the world beyond Singapore, I began a PhD at UCL in London. I chose the United Kingdom partly because DeepMind was here. That motivation may have been naive, but London has nevertheless expanded my horizons—scientifically, culturally, and intellectually. The PhD has not been easy, particularly in its early stages, but I now feel that I am making progress towards many of the aims with which I began it.
During my PhD, I also explored the world of startups. I co-founded Aether Space, a space-logistics company that brokers launches and optimizes rideshare missions, and served as its CTO. We raised 100,000 in funding at a 10 million post-money valuation. Before that, I co-founded Optio Technologies, which developed agentic AI systems for clients in the Middle East and Southeast Asia. Some of the solutions we built helped our clients win prestigious industry awards. I have since stepped back from both companies to focus on research full time, although I expect to build companies again.
There is much more I could say about those experiences. They profoundly shaped how I think about finance, economics, and the relationship between good business and good science. I do not see the two as orthogonal, although their connection is not always obvious. Andy Warhol once said, “Making money is art and working is art and good business is the best art.” He treated commercial enterprise not as a compromise of creative integrity, but as a possible extension of it. I take a similar view of startups and scientific enterprise.
Startups are essential to a vibrant scientific ecosystem. They create places where PhD graduates can continue doing ambitious technical work, generate opportunities throughout the wider economy, and provide test beds for technologies of strategic importance. For nations, they are not merely engines of private wealth; they can also be instruments of economic dynamism and technological resilience.
It is difficult to summarize everything I have learned and experienced since this journey began in 2017. My path has been nonlinear and decidedly non-monotonic. Yet three commitments have steadily moved closer to the centre of my life:
- God
- My family and friends
- My neighbours
Learning how to serve each of them faithfully—and how to be a good steward of the gifts I have been given—has become the deeper journey beneath everything else.