A place to exchange ideas and perspectives, promoting a thriving innovation economy through public policy
641 Discussions

Scaling Knowledge Graphs & Its Applications in Finance, by Joanne Kim

0 0 384
Joanne-Kim.jpg Joanne Kim, Rising Junior - Duke University, Governments, Markets & Trade Intern

The birth of the digital economy has propagated the development of a wealth of innovations and an even larger investment in big data. As the world becomes increasingly interconnected, safely accessing or collecting data and then effectively running data analyses is critical for understanding the market and becoming a leading company on the global stage. In response to the growing call for innovation and greater digital access, Intel has risen to the challenge with the goal of accelerating the ways in which the company enables progress through its products by investing, researching, and building various emerging technologies.

While artificial intelligence, 5G technology, quantum computing, and other innovations are projected to have enormous impacts on the relationship between machines and people, there exists another novel technology which will forever change the way we analyze big data and gather insights: traversing and scaling knowledge graphs. Thinknum Alternative Data and Intel, two thought-leaders in the graph-analytics space, have been actively innovating this technology. Clifton Roberts, the Director of Global Partnerships and Initiatives under Intel’s Governments, Markets, & Trade organization moderated a recent webinar on their innovations.

“I am pleased to join two experts today, who through their respective software and hardware innovation leadership positions, are empowering finance-related organizations with an enhanced capability to process large graphs with billions of edges and move towards even larger graphs with trillions of edges,” opened Roberts. “Such analytics, of course, can drive faster, deeper, and meaningful insights from the data to drive better decision making.”

Nikhil Deshpande, Director of AI Engineering in the Data Platforms Group at Intel, also expressed the promising nature of the technology as he stated it to be “a gamechanger for your business.” Deshpande then went on to discuss the details of Intel’s newest work – Intel® Unified Memory Architecture (UMA) - alongside Marta V. Lopata, Chief Growth Officer at Thinknum, who spoke about KgBase and its projected use cases.

As an alternative web data platform leveraged by leading investment firms and corporations, Thinknum has become an innovative and active player in the data business. Pushing the boundaries of data analysis and visualization, Thinknum launched KgBase - their no-code, collaborative knowledge-graph tool in April of 2020. Knowledge graphs are best known for their strategic role in the development of advanced search engines and recommendation systems, but they also have countless valuable applications in finance, business, research and education. KgBase can show the nature of various relationships and the meaning behind them, making the tool useful for providing critical insights. According to Lopata, a tool such as KgBase can be used to identify risk exposure, help venture capital firms identify good investments, and be used for research and due diligence.

Intel also quickly joined the graphic analytics space after recognizing it to be the next frontier. The company sought engineering solutions that could drive actionable impacts and eventually inform billion-dollar market decisions. Experts, such as Deshpande, immediately tackled the architecture of the traditional hardware and realized it was suboptimal: the original model had limited I/O bandwidth, high latency networks, and a high compute to I/O ratio making it unbalanced. Harnessing the opportunity for innovation, Intel built Intel® UMA – a new model that can capture real-time insights as it traverses through or navigates graphs 1000x faster. Intel® UMA was built without any switches, resulting in extremely high bandwidth, low latency, a balanced compute to I/O ratio, and the capability to process random accesses to account for the unique patterns of graphs. Deshpande excitedly commented, “Just imagine you can predict something because you know something that you were missing before – you know the patterns that you missed before. That’s going to change your investment strategy, that’s going to change how you look at other companies – that’s gold.”

Not only has Intel identified an opportunity for innovation and successfully built a product that will forever change real-time data analytics, but the company has also thought through the positive implications the technology could have for customers, partners, and governments. Predictive analytics – and specifically knowledge-graphs - will be able to play a major role in several spaces: investment insights, fraud prevention, drug discovery, drug repurposing, cybersecurity, government analytics, enterprise AI analytics, and social media insights. Essentially, Intel® UMA could be used to provide critical insights in several systems and situations. For example, considering the current pandemic, Intel® UMA could be used in conjunction with software, such as KgBase, and successfully drive graph analytics real-time to identify key trends, patterns, or relationships related to the virus. Deshpande went on to say, “I really am hopeful that this technology and the graph analytics will eventually help us come up with drug discovery as well as the repurposing of existing drugs for solving these problems [related to COVID-19]. The virus mutation needs to be predicted.”

Ultimately, Intel® UMA will revolutionize the digital economy, as the hardware provides a powerful ability to traverse through large datasets and graphs and provide real-time insights. Its many applications and numerous positive implications will make Intel® UMA an impactful product that can work in conjunction with various software to drive better decision-making. Further, Intel® UMA will surely have an effect not only on business operations, but on society, as it fills the gap and need for quick, deep insights that ultimately add meaning to big data. Roberts concluded, “This increased efficiency to know before others, to extract deeper insights, and to predict with increased confidence – that could be the difference of millions or billions of dollars depending on the size of the assets and portfolio being managed.”

[Joanne Kim is a rising Junior at Duke University interning within the Governments, Markets, and Trade Group at Intel Corporation.  She is dual Public Policy and Psychology Major]