Valley Honors Prof. Thomas Kailath: Educator, Entrepreneur
By Francis C. Assisi
"I belong to the school that holds ideas and exposition to be more important that 'mere' results." Quotation from the preface of Prof. Thomas Kailath’s classic Linear Systems book (1980).
24 February 2006 -- Despite his diminutive size, Professor Thomas Kailath of Stanford University has been straddling the world of academia and industry, mathematics and engineering, like a colossus -- for 45 years.
It’s a record unmatched even by India’s brightest and best students who came to the US in the 1950s, and later. Indeed, at age 71, the Hitachi America Professor of Engineering is still going strong.
Today, as he is being inducted into the Silicon Valley Engineering Hall of Fame, one realizes that the only award missing from the long list honors conferred on him is the Nobel Prize – though the Institute for Scientific Information’s (ISI) database of “Highly Cited Researchers,” has judged him to be “Nobel Class” as a result of his contributions to two related disciplines: Mathematics and Engineering. This ISI classification of scientists as ‘Nobel Class,’ is a provocative reminder to the scientific community that the Royal Swedish Academy cannot always honor all those whose work deserves to be celebrated.
The Silicon Valley Engineering Hall of Fame Award, recognizing area scientists who have demonstrated significant engineering achievements, and provided significant guidance in new and developing fields of engineering, comes at the fag-end of a long and distinguished career for a professor with one of the longest innings at Stanford – 44 years. He is the only Indian-American who is a member of the National Academy of Sciences (NAS), the National Academy of Engineers (NAE), the American Academy of Arts and Sciences and the Indian National Academy of Engineering.
At this point it may be appropriate to consider the overall legacy of Kailath. He has guided 74 doctoral students (eight of them Indians), mentored 40 post-doctoral fellows, co-founded at least six companies, consulted at various levels for US defense agencies, advised the Government of India, and has been honored with at least four doctorates. In addition he was one of the first Indian-Americans to fund a chair in India Studies (in honor of his wife Sarah Kailath) at UC Berkeley.
Kailath has authored, edited and co-authored several books, including "Linear Systems," "Indefinite Quadratic Estimation and Control" and "Linear Estimation". He has held Guggenheim, Churchill and Humboldt fellowships, among others. He served as president of the IEEE Information Theory Society in 1975 and received its Shannon Award in 2000. He has also received honorary degrees from Sweden's Linkoping University, Scotland's Strathclyde University, Spain's University of Carlos III and France's University of Bordeaux.
“It's been a joy to have worked in so many different fields with a stellar collection of more than seventy doctoral students and more than thirty postdoctoral scholars, many of whom are now leading figures in their field s…These efforts have also gained my family and myself a wide range of wonderful friends from around the world” says Kailath.
Beginning as a graduate research assistant at MIT’s Research Laboratory for Electronics (RLE) Information Transmission Group from 1957 to 1961, Kailath’s academic career has been nothing less than spectacular— indeed it is a trajectory that began during his undergraduate years at the College of Engineering, University of Pune.
In his stellar career as an academician, inventor, and an entrepreneur, he is the founder of the Information Systems Laboratory (ISL) at Stanford University (which has played a pivotal technology-transfer role in the growth and success of the Silicon Valley, and boasts numerous highly successful and influential technology companies that have their roots in the research work done in the laboratory). He has founded several industry-leading high-tech companies, including Integrated Systems (now part of WindRiver Systems), Numerical Technologies, Inc, Excess Bandwidth Corporation (now part of Virata Corporation), Cipher State Inc., and Clear Shape Inc.
Most recently, Kailath has been involved in “breaking the 100 nanometer Barrier in Optical Microlithography via Signal Processing.” The company which Kailath co-founded in 1995 -- Numerical Technologies -- makes tools centered on the automated design of what are called Phase Shifting Masks, in which the underlying mathematics is the approximate solution of a nonlinear inverse scattering problem. This work brings several ideas from communication theory into optics, Kailath explains. The main advance is that manufacturers, using current technology and sources (the "0.18 technology" generation), can now comfortably beat the "0.1-micron barrier" in optical microlithography---a very cost-effective acceleration of Moore's famous "law" by several generations.
It was in the 1990s that Kailath began looking into semiconductor manufacturing, which did not seem to be making use of techniques from the "systems side" of electrical engineering. The result for Kailath and his students “has been ten very enjoyable years devoted to the development of solutions for industry, with the not unexpected bonus of new theoretical results along the way.”
Over the past four decades, Kailath's research has spanned a large number of disciplines: emphasizing information theory and communications in the 1960s; linear systems, estimation and control in the 1970s; VLSI design and sensor array signal processing in the 1980s; and applications to semiconductor manufacturing and digital communications in the 1990s. Besides being an educator par excellence, Kailath’s professional success comes from the fact that Information theory has supplied both the architecture and the analytical tools that govern modern digital communication systems.
Watching the growth of Silicon Valley from his Stanford vantage point since 1963 has been “an inspiration and a thrill” for Prof. Kailath. That’s why he admits: “There's a tremendously innovative and competitive atmosphere here. Everywhere there are people who are ready to seize or make opportunities to do something better or faster, or to do something completely new. Almost every company here has spun off or inspired several others.”
And for Kailath, there is much more to a technology company than just pure technology. “Being involved in a high-tech venture has also been a great experience: learning the importance of discipline, product definition, and market analysis; understanding the financial imperatives and learning to live with Wall Street quarterly reports; and gauging the interpersonal dynamics within a company.”
Kailath once said that his idea of teaching and mentoring is to “take an inexperienced student and have him or her cross the threshold as an independent investigator in four or five years.” Another continuing challenge was “to organize the extensive bodies of knowledge that have built up through our research and publish them in textbooks and monographs.”
Sputnik had just happened when Thomas Kailath landed at the MIT campus in Cambridge, Mass. “Not many students came from India in those pre-Sputnik days” recalls Kailath, who became the first student from India to receive a doctorate (1961) in electrical engineering at MIT. But for the insistence of a family friend and his department chair, Professor Aija, Kailath probably would have accepted a civil service job rather than go to MIT.
This is how Prof. Kailath recalls those heady days in the late 1950s at MIT: “It was a great time to be at MIT. The Research Laboratory in Electronics (RLE) had a remarkable group of faculty and a striking collection of visitors. ..I was Jack Wozencraft's first doctoral student, and he steered me towards the study of time-variant multipath channels. It was a hot area at the time and my master's thesis drew the attention of Bob Price and Paul Green at Lincoln Laboratory. I had a tempting offer from Bell Labs to get my PhD in New York City as a paid employee. Fortunately, Bob and Paul persuaded me to stay at MIT by offering me a summer job…That work led to my first journal paper. Although travel to foreign conferences was a rarity for graduate students then, I also presented a paper at the 1960 London Symposium on Information Theory. There I met Professor Norman Abramson of Stanford, who later persuaded me to join Stanford's faculty as an associate professor in 1963.”
It was a time when, as an MIT colleague recalled, “Starting companies, making millions, developing real applications was secondary. There was interest in bringing the theory closer to reality, but it was theory-based. Our role models were relaxed, curious, and had time to reflect.” He adds, the combination of new mathematics, simply stated new problems, and intellectual culture was irresistible to the best students – Kailath being one of them.
However, it was at Stanford and in the heart of Silicon Valley that Kailath bloomed and shone like a star in the engineering firmament.
Let's follow Kailath’s trajectory in his own words: “At Stanford, I worked largely on information and communication theory and stochastic processes. I followed the MIT tradition in which professors worked on their own research while guiding doctoral students along different and separate paths. However the work on feedback communications by my first student led to the need for recursive estimation algorithms that had been developed in control theory.”
Admits Kailath: It was more efficient to let my graduate students teach me this, and I realized that joining forces with them on research was more valuable for everyone. This led to a decade of activity in control and linear systems theory.
In the late 1970s, the potential to blend integrated circuits and systems became evident. I was inspired by the team projects of my solid-state colleagues, who gathered groups of students to study new areas together. This led to larger group efforts on VLSI and sensor array signal processing in the 1980s.
As these new interests developed, I continued mathematical work on building the theory of displacement structure. This was originally inspired by apparently different mathematical and computational questions that appeared in all the areas I had worked in: communications, control, signal processing, computation, and pure mathematics.
In the early 1970s, Kailath was consulting for Systems Control, Inc., which was the first commercial group to conduct systems-oriented studies of civilian and military problems. One of the leaders at the company was his former student Naren Gupta (a 1971 Stanford PhD in aeronautics). Together they discussed going forward on their own because Kailath saw that with Ronald Reagan about to become president, “it was clear that defense was going to expand.” According to Kailath, “I thought we had a good chance to compete for some of the business that would soon be offered. Naren secured our initial contract from Lockheed, and I arranged another through the Office of Naval Research. In 1981, we started Integrated Systems, Inc., in a two-room office in Palo Alto Square with one employee.”
After a few years, Kailath and his colleagues noticed that they “were doing jobs well below our estimated costs.” They realized this was because they had put together a set of control design simulation software packages that greatly sped up the analysis and design phase of big projects. So they bundled those packages together with user-friendly interfaces and offered it as MATRIXx, which was the first integrated higher-level control analysis and design package. It was designed to exploit the strengths of the workstations that were beginning to appear.
Kailath is quick to admit that ISI's growth has been almost entirely due to Naren Gupta. By 1984, ISI had about a million dollars in the bank, but chose to acquire “deeper managerial expertise” for growth. Today, by growth and acquisition, its revenue has shot through the $100 million mark. The company’s key target is embedded microprocessors or microcontrollers. “They're not as visible in workstations or PCs, but they're much more numerous. They're found increasingly in all kinds of devices; from microwave ovens and dishwashers to hand-held phone set s, fax machines, automobiles, and aircraft. These used to be four- or eight-bit devices, and they were relatively easy to program, but the complexity grows exponentially as we go to sixteen- and thirty-two-bit microprocessors. These are now cheap enough to be widely used.”
Fundamental science vs. Engineering applications . . .
After devoting about a decade each to theoretical work in communications, control, and signal processing, Kailath’s work has naturally turned towards “applying these ideas to problems in semiconductor manufacturing and more generally in materials processing.” He observes that “Surprisingly, few ideas in what might be called mathematical engineering have been used in semiconductor manufacturing.”
He explains: “There's a different way of thinking on the device side of electrical engineering. There, they start with physics and invent a system for doing something like rapid thermal annealing. After that, their approach is much more empirical and database-oriented. For various reasons, they don't use the approaches and results of modern control, signal processing, or optimization. So from our viewpoint, many of their procedures are inefficient. Volume is what saves them now, which is why Moore's law works. But the time will soon come when our more analytical approaches will be important.”
Kailath believes there is “a great need and many opportunities for technology transfer.” He notes that in control and signal processing, “maybe 90 percent of what is known in universities today isn't used in industry, except perhaps in aerospace and to a lesser extent in the chemical industry.” However, he admits that technology transfer isn't easy. “It must be done through people--students, postdocs, and research associates. It also takes a different scale of funding from what we are familiar with, at least on the systems side of engineering. It's an issue that deserves more attention, especially from faculty.”
Kailath is well aware that the process of research is always a challenge. For example, he asks, how does one enter new fields such as VLSI design and semiconductor manufacturing from scratch? “It's certainly a test of what we profess to teach. It's also challenging to take an inexperienced student and have him or her cross the threshold as an independent investigator in four or five years. Another important yet difficult challenge is to organize the extensive bodies of knowledge that have built up through our research and publish them in textbooks and monographs.”
Says Kailath, “I hope to spend more time on that now. I also want to encourage the implementation and utilization of those ideas, which is something the progress of technology has made much easier to do now than even a decade ago.”
When students and colleagues met in honor of his 60th birthday (1995), many of them reminisced about their student days with Prof. Kailath at Stanford.
“Interactions with Tom have strengthened my conviction that an academic career can be best enriched by hard-working, ever expanding research horizon, and many others. Perhaps the most important is always presenting your idea in simplest possible language. Tom has provided the most critical encouragement, especially during my years as a junior faculty” said Sun Yuan Kung, Princeton University.
Prof. George (Reji) Varghese of MIT, who subsequently became Kailath’s son-in-law, said: “Professor Kailath has influenced my work and life in many essential ways. It was from him that I first learned what research meant. Certain aspects of his style resonated with my own instincts, and decisively shaped my approach to research. Most striking to me are (and continue to be): his interest in connections, in how things fit together and how they can be reconciled; his attention to what I might call thematic research - narrative constantly updated and elaborated as he and his collaborators uncover and understand more - telling a coherent, logical story of what he knows, where it came from, how it connects to other things, what it is good for; arising from these concerns, his continual flow of questions, looking for answers, clarifications, refinements; guiding the whole process, an aesthetic sense that allows him to extract the good, satisfying answers from among those that are merely correct; his love of books, symbolized by the full shelves that were for me the most impressive aspect of his office, extending from floor to ceiling like great wings behind his chair. All outstanding researchers must have these traits to some degree, but in twenty years of observation, I have not seen anyone to match. "
Said Bradley Dickinson of Princeton: "Working with Tom was invaluable preparation for research and teaching in a broad range of areas in systems and signals. Having done a thesis in linear systems around the time that might be called Tom's "innovations heydays," his recommendation and advice were important factors in my landing a late-opening faculty position at Princeton in 1974. I can't imagine a bigger influence, since Princeton has turned out to be the place where my professional and personal life flourished."
H. V. Jagadish, University of Michigan: "By far the most important skill I learned from Professor Kailath is the ability to think about an unfamiliar problem in an unfamiliar area, and to drill down until good research results were being produced. I did this for the first time when TK and I both embarked on an adventure together to learn about VLSI. While the specific knowledge about VLSI or systolic arrays has not been of much value, the research style I learned from TK's example has been of immense value to me since."
Recalled UCLA professor Vwani Roychowdhury, with whom Kailath co-founded CipherState: “In my first summer (1984) TK told me the following: "There are 168 hours in a week; so what have you been doing?"
And Georgia Tech’s Erik Verriest reminisced: “It may be a bit hoky to say this, but in a way, I always considered Professor Kailath as my "scientific father". I learned from him the dedication to a problem (and solution), how to ask ever more questions, and how to try to present things always as simple as possible. If I picked up a bad habit, it is "collecting books.”
Said R.H.Roy, consultant: “Sensor array signal processing was really the main focus of all those years of toil. Professor Kailath had a significant influence on my work, especially in the support he provided for an area of research that was at the time not in the main stream of his efforts. He clearly had the foresight to see where the technology might be able to go if given the right encouragement, encouragement which he provided. Neither I nor the technology would be where it is today without his support.”
A Personal Note
As for this writer, in my social interactions with Prof. Kailath there is a standard joke that he used to fling at me even as we sometimes discussed things Indian – “so when are you going to write that book?” This was at a time when Kailath was a most prolific author in his field. At the time it was also obvious to me that Kailath was extremely well read about literary and socio-political matters. He also made it a point to keep up with news about the Indo-American community.
Though I spent many years at Stanford as a medical researcher, I did not have the privilege of being Kailath’s student or colleague. He was a good family friend. My fondest memory is of the unlikely Malayalam script, at the entryway to his house on the Stanford campus. It used to give me a pleasant jolt.
And this tribute is my way of honoring him and his family – especially his gracious and generous wife, Sarah Kailath, in whose name he instituted a chair in India Studies at UC Berkeley.