Even before the first modern computers were built, scientists and philosophers pondered the question: what problems can (and cannot) be solved using computers? By the 1970s, the early optimism that computers could solve really hard problems had given way to a realization that a host of these problems seemed equally and extremely difficult - giving rise to a field known as computational complexity. Ashok was a pioneer at the interface of computational complexity, logic and the theory of multiplayer games. His work provided a quantum leap in our understanding of these hardest problems in computing - whether the traveling salesman problem, programs for playing games such as Go and Chess, or for evaluating database queries. As a result of the work of Ashok and colleagues, what seemed like one giant morass of hardness was clinically (and beautifully) dissected into layers of differing hardness. For instance, we learned that computers found the traveling salesman problem easier than playing Go, even though both problems seem intractable. Similarly, he showed that there is a hierarchy of database queries in terms of their complexity of execution, which has important implications for building emergent Big Data applications.
Amazingly, Ashok's early work on complexity turned out to have a bearing on our understanding of the limits of parallel computers - a burgeoning field in the 1980s. What was fascinating about this connection: limits on parallel computing typically arise because the different parts of a parallel computer have trouble getting the data to the right place, far more than any intrinsic limits on computing power. Thanks to Ashok's work, we learned that bottlenecks in communication are intimately connected to bottlenecks in computation, a subtle relationship whose full understanding still defies us.
Next, Ashok turned his attention to the efficient use of memory hierarchies. His colleagues and he devised clever methods by which computers in the real world (with caches and virtual memories) could solve many important problems especially quickly. The key was to shuttle data between the different levels of the memory hierarchy in a principled manner concurrent with data processing, so that the data about to be processed always materialized in the memory layers closest and quickest to the processor. As with much of his scientific output, this body of work resulted in a number of influential patents and published papers that were highly cited.
Ashok combined the rare ability for technical as well as people leadership. He rapidly moved up the ranks at IBM, including a stint at the corporate headquarters, until he headed the fabled computer science department at the Almaden Research Center. Many at IBM (and eventually at other companies) recognized that his keenly analytical mind was very adept at building and leading technically excellent teams working on databases, data mining and search. Certainly, some of the core technology in the database industry and in the web search industry were born under his leadership at Almaden.
After Almaden, Ashok embarked on a new adventure, building enterprise search software at Verity, then the market leader (and now, through a series of acquisitions, a part of HP). By this point in his career Ashok had immersed himself in the challenges of searching and analyzing text data (as opposed to structured data in relational databases, his previous passion while at IBM). After Verity Ashok took his skills to Microsoft, where he was a leader in both the web search and the enterprise search businesses. He focused on novel capabilities for Microsoft's Bing search engine, including the identification of entities such as the movie names, actors and actresses, showtimes, etc in Web pages. This work has found its way into many Microsoft products beyond Bing. For example, when you use Xbox to search for your favorite movies, some of Ashok's innovations are hard at work in the background.
No comments:
Post a Comment