National University Health System (NUHS) delivers better patient care and treatment with AI

Updated: Dec 9, 2021

✦ One of Singapore’s public healthcare groups (known locally as “clusters”) has chosen to use an artificial intelligence (AI) system with real-time streaming capabilities to deliver better patient care and treatment across its network of hospitals, polyclinics, specialist centres, medical centre and academic health science institutions.



National University Health System (NUHS) is one of three public healthcare clusters in Singapore equipped with an integrated academic health system and regional health system that delivers value-driven, innovative and sustainable healthcare. Its network covers an extensive 19 hospitals, polyclinics, specialist centres, medical centre and academic health science institutions.


NUHS building, Featured on Brilliant-Online
Part of the NUHS cluster is the JurongHealth Campus, an integrated healthcare development comprising the Ng Teng Fong General Hospital and the Jurong Community Hospital.

NUHS has built an AI production platform known as Endeavour AI based on an NVIDIA DGX A100 system, a system for AI infrastructure, to become the first healthcare group in Singapore with real-time streaming capabilities to deliver better patient care and treatment, collaborate on biomedical research and transform how illnesses are managed and treated.


Inside view of the NVIDIA DGX A100 system. featured on Brilliant-Online
Inside view of the NVIDIA DGX A100 system

With this newly-launched Endeavour AI platform, NUHS can make real-time predictions on diagnosis, progression of diseases, readmissions, risk of falls, and others. This new system will be integrated with NUHS’ Discovery AI training platform to form a complete training and inference system as part of the group’s digital transformation.


Endeavour AI is a software and hardware stack that features streaming data as well as AI tools running micro services. With the capacity to be able to manage up to 150 projects, Endeavour AI will start off with dozens of projects initially before increasing them.


Among the first projects are those that impact the whole cluster, ranging from predictions on how a patient with a certain condition will fare when admitted to a hospital to analysing magnetic resonance imaging (MRI) images. The projects will involve everything from structured medical data to text-based medical data that form the basis for generating chatbots that are conversational in nature.


Dr Kee Yuan Ngiam, group CTO of NUHS and deputy chief medical informatics officer of National University Hospital (NUH) - featured on Brilliant-Online
Dr Kee Yuan Ngiam, group CTO of NUHS and deputy chief medical informatics officer of National University Hospital (NUH).

Dr Kee Yuan Ngiam, group chief technology officer of NUHS and deputy chief medical informatics officer of National University Hospital (NUH), explains that the organisation is undertaking a digital transformation throughout the cluster, with AI as the centre of its digital transformation.


Dr Ngiam’s team has strategised, planned and built in NVIDIA DGXA100 from day one upon the deployment of Endeavour AI. The organisation will be using it for high speed and large volume inference processed by their AI tools.


For example, for every patient who turns up at its hospitals and polyclinics, every time a doctor clicks, saves or free texts, or when a new lab test results are out, an AI tool runs in the background. All the data gets processed by the AI tools and each tool runs about 100-200 inferences per second. This is done hundreds of times per second throughout the whole cluster at a large volume.


NUHS produces between 20 and 30GB of structured data and text daily, or between 1,800 and 2,500 messages per second for each of its hospitals, which translates to about 10,000 to 15,000 messages per second at peak for all of its cluster. The AI tools need to run quickly in the background to absorb all the data on a day-to-day basis.


“We are building a platform that enables multiple projects to run at a time. We have multiple uses for GPUs, largely in training at this point, but certainly we are well underway in operationalising the production use cas