Skip Ribbon Commands
Skip to main content

NTU EEE alumnus recognised in “Forbes 30 Under 30 Asia”

​​​Every year, outstanding contributions by talented hardworking entrepreneurs and game changers are recognized by various organisations and the “Forbes 30 Under 30” is no different. From innovating in technology and disrupting age-old industries to demonstrating immense talent and dominating the world stage this group of young stars shines in more ways than one. 
Roy.png
Dr Subhrajit Roy

In 2019, one NTU EEE PhD alumnus got the recognition that he truly deserves. Dr Subhrajit Roy was a machine learning scientist at IBM Australia. He had pioneered the use of AI for automatic personalised epileptic seizure forecasting and detection, and his research could help patients with epilepsy to have a wearable, real-time warning system. Dr Roy has authored 31 papers in this field, and his work can change the lives of 65 million epilepsy patients in the world. “It is very inspiring! It feels amazing that our work got recognized on a global scale. It’s an absolute honour!” he said. It has been a year after the event and hence he has had more time to digest the positive consequences. 

One of the significant changes that happened was that Google offered him a position as a research scientist to lead a very interesting project in developing AI technology for analyzing electronic health records. He is currently based in the UK. When asked why he thought Forbes chose him, he said “I think it’s due to the direct practical impact of our research. Forbes typically honours individuals who have directly contributed to society. The system that we developed was one of the first practical and accurate devices for seizure prediction in epileptic patients. It has the potential to make the lives of epileptic patients better by reducing the unpredictability of seizures”. 

The decision to pursue PhD was a deeply considered one with an interesting answer. During his undergraduate studies at Jadavpur University where Dr Roy graduated with Bachelors of Engineering (Electronics and Telecommunication Engineering), he undertook a research project on evolutionary algorithms under Prof Swagatam Das who was a professor at Jadavpur University. Their collaborator on this project was Prof Suganthan who was at NTU EEE. Hence, it was natural for Dr Roy to go to NTU to complete a part of this project and co-write the corresponding research paper that led to an internship under Prof. Suganthan. The month he spent in NTU was one of the most interesting and intellectually stimulating experiences of his entire undergraduate studies. This is the time when his love for research was primarily shaped. 

Hence, as soon as he completed his undergrad, he came to NTU to pursue his PhD to continue on the path of research where he graduated in 2016. Apart from the excellent courses he took at NTU that enhanced his knowledge of the field, Dr Roy worked with under the supervision of Prof Arindam Basu on low-power hardware implementation of neural networks as a part of his PhD thesis. This was very helpful since he used some of the knowledge gained during this time to build the low-power seizure prediction device which was the primary reason behind the Forbes honour. 

Dr Roy’s research area was to develop hardware efficient neural network algorithms. Most artificial neural networks use high-resolution weights and this makes them use more power and memory. This makes them difficult to be used in wearable devices. Hence, to develop the next generation of truly smart wearable devices, they need to decrease the power and memory requirements of neural networks. Dr Roy and Prof Basu developed a new class of learning algorithms to train neural networks with binary weights that showed both state-of-the-art performance and had lower memory and power consumption. That research secured him the role of research scientist with IBM Research Australia. 

At IBM, his job was to develop novel deep learning algorithms for predicting and detecting epileptic seizures from electroencephalogram (EEG) recorded from the brain and other physiological signals. His team worked on developing a wearable seizure prediction system for epileptic patients. This system is based on machine learning algorithms, long-term seizure data, and an ultra-low power neuromorphic processor. They provided proof-of-concept for this device, demonstrating that it is accurate and fully automated, as well as patient-specific and able to be tuned to the needs of individual patients. 

The algorithms implemented in this system enable the patient to make instantaneous adjustments to the balance between device sensitivity and alarm time. Compared to equivalent random seizure predictors currently available, Dr Roy’s system offered a 42% improvement in performance. 

Currently, Dr Roy have joined Google as a research scientist and moved to London, UK. At Google, he work as a part of Google Health, and his team designs deep neural networks that can predict adverse events in hospitals.

Published on 6 June 2020​​​
Not sure which programme to go for? Use our programme finder
Loading header/footer ...