A Personal Journey in Decision Sciences and Engineering Systems

The author was born on March 27, 1945, in the City of New York where his China-born parents were sent by the then nationalist Chiang Kai-Shek government to negotiate with the United States government for a post-World War II Marshall Plan for China; the Plan was a generous American initiative enacted in 1948 to provide U. S. financial aid to the war-torn allies of World War II. Although it was agreed that China would be a part of the Marshall Plan, unfortunately it was not to be since China’s communist party, under the leadership of Mao Zedong, defeated Chiang’s government in 1949. Later, the author’s parents, together with his four siblings and two grandparents, emigrated to Sao Paulo, Brasil, in 1950, followed by a subsequent emigration to New York in 1959.

As indicated in Table 1, the author began his formal education at two English boarding schools; first at Escola Britannica in Sao Paulo, Brasil, and then at the Stony Brook School in Stony Brook, New York, USA. He then enrolled in electrical engineering at Rensselaer Polytechnic Institute and graduated with a Bachelor’s degree in 1966. He subsequently expanded his engineering systems focus to encompass decision making with a 1972 doctoral degree in operations research at the Massachusetts Institute of Technology.

Coincident with his educational endeavors, the author obtained employment at four decision sciences and engineering systems (DSES) related companies - Bell Telephone Laboratories, The Rand Corporation, Urban Systems Research and Engineering, and Structured Decisions Corporation. These four early jobs helped sharpen his academic interests in research and education, after which he sought and obtained 48 years of employment at two well-regarded universities: 30 years at Rensselaer Polytechnic Institute and 18 years at the University of Miami. Moreover, his tenure at these academies were, at RPI, as 1) the founding chair of a nationally-ranked department of decision sciences and engineering systems (alternatively - and sometimes - referred to as department of industrial and systems engineering); and 2) Yamada Professor, named after the founder of Yamada Corporation; and then, at UM, as 3) Distinguished Professor; 4) Dean of Engineering; and 5) University Ombudsperson.

In parallel with the two professorial positions, the author was elected to the covetous US National Academy of Engineering (NAE) in 2001 and later served as Chair of NAE Section 8 and as the NAE International Secretary. (Table 2)

Section 3 identifies the DSES-related pragmatic attributes that have combined to provide the author with a valuable education and a productive employment. Data analysis, data modeling, operations research [4, 5], optimization and risk, and big data analytics are decision sciences related attributes, while software systems, hardware systems, communication systems, internet of things [1‐3], and intelligent systems are engineering systems related attributes.

Thus, a Covid-19 analysis ranging from the epidemic, pandemic, and endemic phases is considered; the fourth industrial revolution focusing on real-time customization is considered; and the application of technology to healthcare is likewise considered. Additionally, the convergence to real-time decision making; the integration of internet of things, real-time decision making, and artificial intelligence is considered in a combined manner; and autonomous vehicles are regarded as the sputnik of servgoods (which is a combination of a service and a good). (Table 3)

In regard to investments, the author has likewise employed his DSES-related skills to make DSES-related investments or stock purchases; the purchases have not only been profitable but have also contributed to the author’s personal wealth. While some of his stocks have yielded minimal growth or have sometimes regressed to a zero value, it should be noted that three of his DSES-related investments - namely, in MSFT, AAPL and NVDA - have yielded high positive returns. Indeed, the three stocks identified in Table 4 are now the author’s highest valued technology investments, each is currently market valued at over a three trillion dollars! To be truthful, although the author was able to accumulate a significant amount of MSFT and AAPL in the 1990s, he was initially too stretched (i.e., too over-committed) to purchase more shares of both MSFT and AAPL. On the other hand, because he had early focus on NVDA, starting in 2018 and at an initial cost of $3.18 a share, he was able to accumulate a larger amount of NVDA shares.

Prospectively, the author foresees a continuation of DSES-related investments. However, in the next phase of such investments, there must be larger data centers with perhaps novel processing units, or a combination of processing units, including central processing units (CPUs), intelligence processing units (IPUs) and graphical processing units (GPUs). More importantly, the large AI data centers must be powered by more efficient and effective energy sources, including wind, solar, ocean waves, and perhaps even nuclear fusion or enhanced geothermal systems (EGS).

There are a range of challenges in the management of AI applications. First, there is concern about data quality. How can we ensure that the volume and variety of data feeding AI are accurate, so that the AI-provided answers are also accurate? Security challenges arise when applying governance practices to large amounts of data, including maintaining a clear “chain of custody and compliance”. Training performance and inference performance are related concerns; to return accurate results. AI models require training, which in turn requires fast and accurate processing for large data sets. Performance for inference means getting outputs as efficiently as possible to applications and users. Inference performance is especially critical for real-time situations, like self-driving vehicles. Indeed, the sheer volume, velocity and variety of data consumed and created by AI will always present problems of scale.

In particular, four areas are ripe for efficiency improvements. First, operational efficiency, implying that fewer people must manage larger amounts of data. Second, energy efficiency, allowing for storage growth within an existing power footprint and at a reduced operating cost. Third, performance efficiency, implying high throughput and low latency for both training and inference of AI applications. Fourth, scale efficiency, allowing organizations to quickly and non-disruptively scale up, adding storage devices to existing arrays, or scale out, adding networked storage nodes, to meet the growing demands of AI data and workloads. Efficient scaling must occur in a transparent manner, eliminating any costly downtime or performance impact.

As a consequence, modern storage platforms seamlessly tie storage systems together to provide scale efficiency while increasing operational, energy and performance efficiency.