JEDDAH (Rahnuma): Supercomputers offer incredibly powerful processing capabilities but the high cost and a need for skilled operators means that they are mainly used by government organizations and other large institutions and corporations.
UnitX, a startup based at King Abdullah University of Science and Technology (KAUST), aims to change this by offering on-demand access to the powerful technology, making it easily accessible and affordable to a much wider range of businesses. The company, which was founded in December 2017, provides a cloud-like online platform that allows users to harness the power of 14 supercomputers, three of which are located in Saudi Arabia.
“In the GCC and the Kingdom especially, there is a lot of buzz around artificial intelligence (AI), machine learning and simulations,” said UnitX CEO Kiran Narayanan. “Everybody talks about this but what they don’t talk about is that there’s a huge skills gap in the industry when it comes to data engineers and data scientists.
“Also, the companies that do have this talent on board, they find that their engineers and scientists spend more than 50 percent of their time dealing with software and hardware infrastructure issues rather than focusing on business outcomes. And so there’s a problem of skills and there’s a problem of infrastructure management, because of which companies lose productivity and lose money.
“There’s an access barrier as well: Large institutions such as government bodies have supercomputers but a regular company or customer, an end user, can’t get access to them because the supercomputer centers have security protocols and they don’t trust them. So there is a need for (supercomputers) and there is a supply of them but the two don’t meet. This is the gap we bridge, the problem we solve.
Supercomputers are already being used in the Kingdom, Narayanan said.
“Companies such as (Saudi) Aramco and universities such as KAUST have them,” he added. “What we want to do is democratize supercomputing, this technology that is very fast at solving huge problems such as big-data analytics, machine learning and AI. We want to hand this power to regular industries, not just the largest corporations.”
A supercomputer costs anywhere between $100 million to $250 million, said Narayanan, plus an additional $5-$6 million a year for the power and the skilled team that is needed to operate and maintain it. As a result, businesses face two main constraints when they want to make use of the technology.
“One is the high capital expense and the high level of skill required to own and operate these machines, and managing the infrastructure requires teams of IT people,” he explained. “So the key to democratizing this is to break down these barriers of skill, cost and management issues.
“(UnitX) has a software platform that makes supercomputing as easy as using Netflix. Just as you log in to a Netflix account through a web browser, you log in through the UnitX platform and, just as Netflix (provides) movies, the Unit X platform (provides) applications for big data analytics, AI and machine learning, and simulations that can be deployed onto supercomputers connected to the platform.”