Business intelligence, financial, copyright Rikkyal via Creative Market
Wholesale Chief Digital Officer, Nomura
Merging Science and Finance
1996 was a promising year for the finance industry. Technology paved the way to a new and exciting field known as financial engineering. A field that allowed engineers to apply technical methodologies that result in unique financial products and strategies, essentially merging science and finance. Jezri Mohideen was among the first to capitalize on this development and has been a trailblazer ever since.
Mohideen, a seasoned engineer and science-minded executive, has deep domain expertise in his field and is fascinated by the art of the possible. He now serves as Wholesale Chief Digital Officer at Nomura, a financial holding company headquartered in Japan, where he is a driver of exponential technologies in the world of finance.
But what caused Mohideen to turn to the financial services industry, and how are his contributions making a difference? What technologies is he paying attention to, and what knowledge does he have to share with us?
Mohideen was educated at the prestigious Imperial College in London and MIT, receiving a Ph.D. in process systems engineering. “During my college years, I was fascinated by the Oil & Gas industry due to my interest in computational fluid dynamics and multiphase flow. After interning at Royal Dutch Shell I homed in on the disruptive side of the petrochemical business,” Mohideen tells Cognitive Times. “This cultivated my interest to do a Ph.D. in process systems engineering where the focus was on modeling chemical processes as dynamical systems operating under uncertainty. This took me into the exciting area of non-linear dynamic optimization and stochastic calculus.”
A clear enthusiast of technology and its application to the real world, Mohideen came to realize that the unique challenges of forecasting markets and the potential of applying economist Harry Markowitz’s risk-reward model were too attractive to pass up. “My passion for engineering and real-time problem solving is what pushed me into finance,” says Mohideen.
AI In Finance
According to Mohideen, the financial industry can benefit significantly from AI. “Given the high volume and breadth of data, accurate historical records, and the quantitative nature of the finance world, few industries are better suited for artificial intelligence,” he says. In his eyes AI technologies such as machine learning are paving the way to use cases in financial services like never before. The increasing accessibility of computing power and ML tools will enable this trend to continue.
“Today, machine learning has come to play an integral role in many parts of the financial ecosystem, from systematic trading, sentiment analysis, recommendation engines, predicting credit defaults, to portfolio optimization and risk management,” Mohideen tells Cognitive Times. But in his view, there is a substantial amount of data in the finance industry that has not been processed or understood. Fortunately, the emergence of AI techniques from other sectors has begun to enter the financial services industry, allowing experts to make use of potential sources in both private and public datasets.
Mohideen explains further that financial product pricing and derivatives that calculate risk sensitivity parameters are emerging areas of significance when it comes to AI in finance. “Simply put, the application of AI techniques in this field makes it possible to perform virtually any manner of risk analysis on even highly complex instruments with ‘shadow greeks’, ” he says. “AI offers the potential for pricing derivatives and managing risk on steroids.” To Mohideen, the idea of developing an assembly line that would facilitate the creation, pricing, and risk management of financial derivatives is an opportunity worth looking forward to.
AI In Trading
Prior to his role at Nomura, Mohideen was one of the early adopters of AI in trading at Brevan Howard. But he claims AI-assisted trading has evolved significantly since then and breaks down some of the fundamental points behind his argument.
According to Mohideen, implementing advanced ML techniques in algorithmic trading approaches to expedite trading decisions is common today. Although the majority of financial institutions and hedge funds do not directly make their AI tactics to trading known, they agree that ML and deep learning both play a crucial role in facilitating real-time trading decisions. For instance, robo-advisors represent intelligent algorithms that position an investment portfolio to meet a user’s goals and risk tolerance. Even “data-driven knowledge solutions” are deployed on a large scale so that industries can better understand customer behavior, habits, and preferences.
When it comes to trading, Mohideen stresses the level of frequency trading and where it stands today with the integration of AI systems. “Currently, the application of AI to drive trading decisions is far more established in the high-frequency domain than in mid to lower-frequency trading. One of the reasons for this is simply a reflection of the very large quantities of data required to train the current generation of machine learning models. In the original context in which many of the most successful AI techniques were developed e.g. image recognition, vast data sets, and static data generation processes (a cat in 2020 looks intrinsically the same as a cat in 1950) enabled brute force data-driven learning approaches to be spectacularly successful,” Mohideen tells Cognitive Times. “In the high-frequency trading space, a year’s worth of historical tick data can provide sufficient data points for such data-intensive techniques to be applied effectively. However, this comes at the expense of a limited capacity to deploy capital, as market impact effectively restricts the sizes of trades that may be executed.”
But Mohideen notes that lower trading frequencies create more challenges. “At the lower trading frequencies of hours, days or weeks at which the larger amounts of capital more typically traded by hedge funds and institutional investors are deployed, the relatively smaller quantities of historical data coupled with the constantly evolving nature of the market environment over longer time frames present a significant challenge to the successful application of pure data-driven approaches to extracting alpha (i.e. over-market return on investment). In this domain, it is the combination of the application of AI techniques to extract meaningful signals and insights from much broader cross sectional data sets – not just market data, but fundamental data and alternative data – coupled with the domain expertise of traders or portfolio managers in crafting trading strategies around those signals that is likely to prove most successful,” he says.
However, Mohideen believes that we are not to the point where AI has become the driving force behind trading initiatives. He explains it, rather, with the paradigm of Kasparov’s Centaur. The way Mohideen uses it, the term centaur refers to using human-machine teaming to leverage information that improves or complements decision-making. Or as he puts it, “where the machine is used in partnership with the human trader.” To Mohideen, this partnership model harnesses the potential of AI-oriented solutions to deliver more value by making up for some of the inherent weaknesses that human decision makers display, such as behavioral or cognitive biases. Also, these AI-driven solutions extract unbiased, definite, and meaningful signals and insights from complex relationships in raw data that would not be visible to a human observer.
We can safely say that AI is one of the most integral technologies of our time. But the topic of whether it can truly predict and ‘learn’ financial markets is far from decided. There are even larger questions about how AI might disrupt the financial industry and how financial services can harness its full potential, not only to mine “alpha” but also to contribute stability. Despite the successes AI technologies have already demonstrated, does the market remain inherently unpredictable and unlearnable by AI?
According to Mohideen, AI provides its users with a competitive edge in making profitable predictions. In essence, trading and investing come down to understanding, forecasting, and trading causality. The presence of algorithmic intricacies in financial markets makes it near impossible for professionals to capture them by using multilinear models. But AI can offer an advantage. “Financial markets are profoundly complex dynamical systems, with deep algorithmic subtleties that are hardly likely to be captured by invoking multi-linear models. So at least from an intuitive point of view, AI should offer a hedge,” he tells Cognitive Times.
Computing offers several possibilities, and Mohideen connects this advancement to Bennett’s logical depth which is a measure of complexity for individual strings. Logical depth involves computational resources taken to calculate the results of a program of minimal length. It measures computational content instead of informational content. So, if you apply the concept of Bennett’s logical depth, you will discover that only intensive computing structures can simulate systems with deep causality.
“AI is computational learning, and ultimately a machine is capable of learning on specified problems faster than a human brain. Ultimately learning is compressing complexity by accepting a given amount of uncertainty,” says Mohideen. As for examples, Mohideen provides a few: “AI solutions will outperform humans on very well-specified problems with a limited scope, in particular enabling humans to explore all the unseen frontiers that remain out of reach because of our own natural computing limitations. AI solutions need to consume very large quantities of high quality data to capture these algorithmic subtleties, given the current state of the art. AI solutions are always specific; in the same way that you would not expect Alpha Zero to be used to drive your car, you cannot expect that an AI solution that has been designed to trade government bonds will be useful to build a general risk allocation solution.”
Venturing into Crypto and Other Exponential Technologies
An engineer at heart, Mohideen is not just an expert in AI, but he has also helped lead Nomura’s efforts in cryptocurrency. In June 2020, Nomura, together with companies Ledger and CoinShares, formed Komainu. The joint venture, designed by finance and security leaders, is a digital asset custodian business for financial institutions. The Komainu platform was designed to overcome barriers that hindered institutional investment in digital assets by offering an infrastructure and operational framework for the financial industry.
“With the ten largest digital assets alone reaching a total combined market capitalization of $500bn+, institutional investors are looking to incorporate digital assets into their investment strategies. Komainu will address the need of these investors for decentralized finance that is institutionalized, regulatorily compliant and secure. Through this joint venture, we successfully bridged the gap between legacy finance and emerging technology, establishing Komainu as a regulated and secure digital asset custody solution tailored to the needs of institutional clients. This will act as a foundational pillar to Nomura’s digital asset strategy, enabling the firm to further progress with its initiatives across the full digital asset value chain (e.g. Asset tokenization projects, stable coin/CBDC’s, etc.), thereby setting out to become one of the leaders in the institutional digital asset industry,” Mohideen tells Cognitive Times.
AI is the next big thing in virtually every industry and the cryptocurrency sector is no different. Organizations are keen to cash in on AI because of its immense potential and its long list of successes across multiple industries. The question is: just how will AI meet cryptocurrency markets and digital ledger technologies (DLT)? What will these hybrid solutions look like?
Mohideen believes a coming convergence will lead us to an exciting world of cryptocurrencies and asset tokenization. As he puts it, “where crypto currency has definitely taken the world by storm, cryptographic tokens are comparatively more nascent. Tokens can represent underlying securities, physical assets, cash flows, and utilities. AI provides the means to process these large data sets in block-chain in order to predict the movement of these tokenized assets.”
Mohideen is also quick to point out the importance of bitcoin itself, as it makes up a significant share of the entire cryptocurrency ecosystem. He believes crypto can and will hedge away the risk of fiat currency debasement. On the other hand, “digital ledger technologies could significantly change the finance industry by reducing transaction costs and decreasing the time taken for settlement, to improve audit accountability,” he says.
But to Mohideen, the questions surrounding cryptocurrency and decentralized finance are the most exciting. He finds the situation to be fairly straight forward: once regulatory concerns are addressed and overcome, there will be a massive adoption of the technology. Moreover, he describes financial markets as a social technology, so if changes are brought to a system, it redefines how people interact.
“Let’s call a script a collection of strings (words, numbers) that means something. Essentially, finance is processing scripts: from accounting entries to highly complex derivatives formulae. It is all scripts and by definition they are food for machines,” says Mohideen. “But the crypto-economy is first and foremost a technology to exchange these scripts in a safe and decentralised manner. That is the very essence of smart contracts. Thus, it is easy to understand that the future of finance could live in a crypto-economy of some sort. Sooner rather than later, it will be feasible to create highly customized solutions ‘on the fly’ for any sort of client (SME included), tokenize these solutions – meaning morph them into interoperable programs that are digested by a protocol – and distribute them at a lower cost and at an unrivalled scale. These solutions then become part of the knowledge base and it is possible to reuse them and iterate. Indeed the very same technology combination of AI + Finance + Crypto can be applied to many old classical problems such as risk allocation. The essence of investing is about risk vs. reward, and non-professional users often struggle to understand and quantify all these risks. It would be a good thing if we could make various different risks look like very well-specified objects that can be expressed in terms of small programs.”
To Mohideen, there is a lot to be excited about in the world of finance, an industry well suited for a unparalleled merger with artificial intelligence and computational science. At Cognitive Times, we couldn’t agree more!