The IHPC team comprising of Dr Liu Yong, Dr Rick Goh, Dr Zhang Weiwen, and Mr Wang Long won the first prize at the inaugural National Supercomputing Centre (NSCC) Hackathon 2016 in the finals held on 6-7 Aug 16.
The hackathon aims to explore business problems that can be solved using a supercomputer, and attracted nineteen entries from local and international teams. IHPC's entry focused on ModStore, a large-scale high-performance-computing-based collaborative machine learning platform that empowers the data science community with diverse competency levels to model build, test, deploy and monetise data science models. The innovative business model presented by the team is to match the supply and demand. Supply comes from scientists who develop data science models (e.g., A*STAR, universities), while demand can hail from various business entities ranging from manufacturing, services and digital economy, healthcare to urban systems. The judges commended the team for the innovation in building data science models off a typical internet browser, where the compute-intensive models run on supercomputers. This hides the complexity from the community at large when using large high performance computing infrastructure, and to focus on data science that can be easily deployed and used in web and mobile apps.
The team won 500,000 CPU-core hours in NSCC, S$5,000 in prize money, as well as a Macbook, iPad Pro, and the latest NVIDIA GPU card. The team intends to use the 500,000 CPU-core hours for our collaboration with Dextra which is Singapore's largest data innovation challenge organiser that also has a private data marketplace with 100+ datasets from private and public entities, and 5,000+ developers. ModStore will be the platform used by these developers in various hackathons and competitions. Also, IHPC is working with MOE & NRF on a Data Challenge later this year where up to 10,000 students (secondary, JC, ITE, polytechnic) will use ModStore to do data science using the National Science Experiment (NSE) data that the students have collected. We are also working with Singapore Land Authority to incorporate OneMap 2.0 geospatial information into ModStore to allow better visualisation of location-based information. These efforts are aligned with Singapore's Smart Nation agenda and ambition.