Scientists power up research with Singtel AI cloud

Image: Asst Prof Wilson Goh (right) and Yoo Sehwan, who used RE:AI to develop AI tools for biomedical applications.

To build smarter and healthier cities, scientists first need to understand how air moves through them.

At the School of Mechanical and Aerospace Engineering, Asst Prof Li Hongying is using advanced simulations to study how airflow, heat and particles move through urban environments, accounting for how wind funnels between buildings, how heat builds in dense areas, and how pollutants travel through the air.

The work could give cities sharper insights into air pollution, urban heat and ventilation, helping planners test different urban scenarios and design healthier, more sustainable spaces. But such simulations are highly complex, and until recently, a single run using conventional processors could take over a day.

The same need for faster, more powerful computing applies in healthcare. At the Lee Kong Chian School of Medicine, Asst Prof Wilson Goh is developing AI tools to analyse high-resolution pathology images and make tissue assessment faster and more reliable.

Doctors often need to make decisions based on images that can be difficult and time-consuming to interpret. By processing large collections of medical imaging data, Prof Goh's work aims to provide doctors with more timely information, which may in turn support better patient outcomes.

Profs Li and Goh are among 19 NTU research teams - spanning disciplines from urban sustainability and healthcare to robotics and seismology - now drawing on Singtel Digital InfraCo's AI cloud service, RE:AI. Powered by Nvidia H200 GPUs, the service has compressed simulations that once took more than a day on conventional CPU-based systems into about 10 minutes.

That capacity is reshaping work across the University. Assoc Prof Soujanya Poria from the School of Electrical and Electronic Engineering, trains AI models to help robots respond better to real-world situations and make sounder decisions in settings such as factories, laboratories and service environments.

Image: EOS and ASE Asst Prof Luca (2nd from left) and his researchers used RE:AI to analyse data from past earthquakes and to run realistic simulation models.

At the Earth Observatory of Singapore and the Asian School of the Environment, Asst Prof Luca Dal Zilio uses PyQuake3D, an in-house computational framework, to analyse data from past earthquakes and run larger, more realistic simulations of earthquake and fault behaviour. This could help scientists develop better models of seismic hazards in the future.

Professor Luke Ong, Vice President for AI and Digital Economy at NTU, said: "NTU is accelerating AI adoption and development across disciplines. This includes building up the University's AI compute capabilities available to our research community. RE:AI provides on-demand GPU capacity for computationally intensive work in areas such as climate change, geological hazards, robotics and biomedical engineering, where AI is increasingly central to scientific discovery and real-world impact. By cutting weeks or months from such research, RE:AI can help our scientists shorten the path from idea to experiment."

Mr Bill Chang, Chief Executive Officer, Digital InfraCo, Singtel, said the platform supports Singapore's Research, Innovation and Enterprise 2030 roadmap by giving researchers access to secure digital infrastructure and sovereign AI capabilities: "By equipping NTU with our sovereign AI cloud and on-demand GPU-as-a-Service, we are strengthening its R&D capabilities and helping position Singapore as a leading destination for world-class research talent."