Profiles in OR/MS: Brenda Dietrich
Senior Research Manager,
B.S. UNC-Chapel Hill,
PhD Cornell, Operations Research and Industrial Engineering
Questions & Answers
Q. Tell me about your academic and professional background, and how you arrived at your current job as head of the Optimization Center at the IBM T. J. Watson Research Center?
A. My undergraduate degree is in Math, with most of the course work in algebra and number theory. I studied math because I liked it. I started graduate school at Cornell in the Math department. I took my first O.R. courses, game theory and combinatorial optimization as part of the “outside minor” required in the Math department. These courses captured the parts of math I liked most – the linear algebra, the geometry, and the discrete math, so I switched departments, to Operations Research, after my first year.
I worked for IBM during the summers while at Cornell, in three different departments, doing graphics programming, optimization algorithms, and manufacturing models. In 1984 I joined an O.R. group in IBM research that worked on optimization and simulation applications for IBM manufacturing. I did not seriously consider an academic career. In 1992 I became manager of that group, in 1995 I became a senior manager, and in 1997 I combined a logistics applications group with a network applications group and a more theoretical group and became the manager of the optimization center.
Q. What was the main reason for merging these three OR groups into one Optimization Center?
A. I asked for the change. I wanted to be involved in a larger range of activities, and I felt that we’d achieve some synergy by combining the applications groups with the theoretical group.
Q. I recently visited the IBM Research website and was impressed with the broad range of operations research topics currently under study at IBM. Can you tell me more about the research that is currently going on in each of these different areas?
A. Most of the O.R. work in IBM Research is done in two departments.
One of the departments, Enterprise Systems – focuses on enterprise applications, and uses a variety of techniques – optimization, simulation, expert systems, etc.
My group uses primarily optimization, and works on core solvers, such as the Optimization Solutions and Library (OSL), and industry applications.
The two groups often work together, and we have an O.R. professional interest group (web site above) that includes most of the O.R. researchers in IBM, as well as some of the IBM consultants.
Q. What are some of the industry applications that the Optimization Center is currently working on?
A. We work with airlines, with retail companies, and with transportation companies. We have also build optimization applications for IBM manufacturing and for IBM Global Services. The applications we build for IBM are also made available to some IBM customers.
Solution techniques vary, but usually involve either solving large integer programs or building special purpose heuristics that exploit the underlying problem structure.
Q. What are some of the challenges you face in trying to solve these problems for industry?
A. One of the most interesting challenges we face, is that as our customers see the benefits of using optimization-based applications for one business problem, they find many other similar problems they want solved. We try to find a balance between general-purpose optimization engines that can be customized for a class of applications and a highly specialized application, that works really well on a specific problem, such as airline crew optimization.
Q. What are your main job responsibilities as Senior Manager of the Optimization Center?
A. My responsibilities include defining the long term strategy and direction of the group, communicating that strategy up, down, and sideways, identifying projects and partners that are consistent with that strategy, and recruiting, managing, and developing the researchers that will execute that strategy.
Q. What are some of the daily challenges you face in your job?
A. What are some of the ways in which you (and IBM) help to foster the continual development of the researchers that you manage?
Q. What are some of the ways in which you (and IBM) help to foster the continual development of the researchers that you manage?
A. We try to hire the best people. We maintain ties with academia and keep our researcher current in their technical fields through summer visitor programs, seminar series, and campus visits.
IBM has many internal classes to strengthen and reinforce business skills, software skills, and communication skills. We also have mentoring programs and career planning processes.
Our marketing teams and consulting groups are a great source of information about real needs of our customers for optimization applications. Many projects are done by teams, which are usually led by a seasoned researcher. New employees learn by working with experienced employees.
Q. In an article from Think Research magazine (Number 2, 1999), you were quoted as saying, "I want to be working on important hard problems. Problems that have the potential to change the way business is done." What was the most memorable "important hard problem" that you have worked on? How did it affect the way business was done?
A. The most important problem I’ve worked on, and the project that I had the most fun on, was developing software that is used by IBM manufacturing planners to figure out “what they can build with the parts they’ve got.”
When I began working on this problem, the planners were faced with shortages of a few key components, and these components were used in many different combinations in many different IBM products. Planning at each plant was a monthly process that took many days, and was done using calculators and the output of MRP (material requirements planning) runs. The goal was to compute feasible production schedules that made the “best” use of the constrained parts.
The problem is easily formulated as a linear program – but it’s typically has several hundred thousand variables and constraints, and in 1989 most LP solvers could only handle 32,000 variables. We developed fast heuristics to get good feasible solutions, and LP-reduction methods to fit the active constraints and participating variables in the 32,000 limit, to solve the basic implosion (opposite of MRP explosion) problem. With this tool, planners could determine feasible production plans in minutes instead of days.
We worked with IBM plants to deploy the software, and then began a long series of enhancements and improvements that have culminated in the Production Resource Manager, a tool that is used in many IBM manufacturing facilities, and is sold to several IBM customers. In part through the use of PRM, IBM moved from a monthly corporate-wide planning cycle, to weekly or daily planning at each of the manufacturing sites.
This project involved a great team, interaction with many parts of IBM, and a nice mix of the practical with the mathematical.
Q. What are the most valuable technical skills that you believe are needed to be successful in the OR/MS industry?
A. A broad understanding of a spectrum of OR techniques – optimization, simulation, stochastics, statistics, queueing, and,more importantly, and understanding of what techniques work for which business problems. Recognizing an LP or a queueing network when it is embedded in a larger business process.
Some level of programming skill is necessary, because at minimum, the OR practitioner will have to prototype algorithms.
Q. What are some of the most valuable non-technical skills that you believe are needed to be successful in the OR/MS industry?
A. Good communication skills, especially listening, and being able to talk math without jargon or notation, are important. Some general business knowledge, and domain knowledge in some industry are very helpful. An ability to explain highly technical concepts to a non-technical audience, and a sense of humor also help.
Q. In what ways do you continue to expand your knowledge of new technologies and techniques in OR/MS?
A. I read a lot of technical papers – some written by my group, some that I referee for journals, and some that catch my interest when I’m flipping through abstracts on line. I try to attend technical talks at conferences, and attend the seminar series we hold here at IBM. I also like to read trade press – articles about how companies are using technology.
Q. What do you find most rewarding about your career in OR/MS?
A. I like to solve real problems, or at least play a role in getting problems solved—and I like to be able to prove things.
Q. What advice do you have for those just starting out in a career in OR/MS?
A. Learn as much about the math as you possibly can. Then learn about applications of the math. There are usually several ways to attack any problem, so you have to keep a reasonably open mind about which method to use. Get good at programming.
Q. What do you predict the future has in store for the field of OR/MS and for OR/MS practitioners?
A. I think this is a really exciting time to be working on business applications of optimization. Computing power is great and getting better every year. New architectures and increased computing power lead to new breakthroughs in algorithms. The data is available on line and there are programming tools to ease the application development. Companies appreciate the business value of optimization and are investing in applications – and end users are becoming more computer literate and easier to satisfy.