Oktay Günlük
http://www.research.ibm.com/people/o/oktay/
Short Bio:
I did my undergraduate and M.S. in Istanbul at Bogazici
University, Dept.
of IE.
I received my Ph.D. from Columbia University, Dept. of IEOR
where I worked with Dan Bienstock.
After Columbia, I spent a year at CORE (Belgium) as a postdoctoral fellow and two years at Cornell
University, School
of ORIE as a visiting scientist.
In 1997 I joined the OR department at AT&T Labs and in 2000 I moved
to the Math Sciences Department at IBM.
General Research Interests:
Mixed-integer programming, combinatorial optimization,
multicommodity flows. Modeling, optimization and computation, especially
applied to logistics, routing, workforce scheduling and network design.
Professional Life:
My professional life at the Math Department at IBM is a
mix of academic research and consulting projects. In terms of project
work, we are typically approached after a big customer (including
governments) has shown interest in an optimization application and
"real" consultants have seen an opportunity for heavy duty optimization
techniques. Before we go visit the customer we usually have email/phone
conversations to get a good idea about the optimization problem and
based on that we often do some reading to learn the issues related with
the application. For example a couple of years ago we did a
routing/scheduling project for the Port of Singapore and before we visit
the customer we found a few very useful papers that helped us
understand the general problem and existing solution approaches. The
particular problem that the customer needs to solve is always different
from what one finds in academic papers. Depending on the agreement with
the customer, we are sometimes able to publish papers describing the
problem and our solution approach. However, this is not always the case
and after the Singapore project, we could not even talk about it in
conferences. In a different project, we helped a high-end limousine
company "continually" optimize its operations. What we did was a nice
application of column generation to solve a set cover problem where
columns are daily routes of cars and items to be covered are the
customers that need transport. In this project, we were able to write
and publish a paper without using the name of the customer.
Even though there are some opportunities to publish
papers based on projects, these are usually applied papers with limited
new theory. The second half of my professional life at IBM involves
doing academic research on theoretical problems. I have done a lot of
work on generating cutting panes for mixed-integer programs. For
example, one recent topic of interest is generation of cutting planes
using lattice-free polyhedra. Even though it sounds complicated, the
underlying idea is very basic and a lot of known cutting-planes actually
of this type. In a recent paper with Sanjeeb Dash (a colleague at IBM)
and Santanu Dey (Prof. at Georgia Tech) we showed a very interesting
relationship between cuts from 2-dimensional lattice-free sets and cuts
from two-term asymmetric disjunctions. I got interested in this topic
more than a year ago and wrote a paper on a related topic with Ricardo
Fukasawa who was a post-doc at IBM last year and is a Prof. at Waterloo
now. Later I invited Santanu Dey to visit IBM for a few days to discuss
related problems and Sanjeeb Dash also got interested and we started
working on it together. In terms of academic research, we have full
academic freedom at IBM and can work on problems that we like. If there
is project work, however, it takes priority over research and we need to
park or slow down research activity. This has happened to me several
times.
In addition to project work and research, we also do some
service to the academic/professional community. In the past I have
helped organize workshops, conferences, did refereeing and editorial
work for academic journals and participated in NSF panels. Currently, I
am the vice chair for Integer Programming in the Informs Optimization
Society and I am putting together the Integer Programming session in the
next two Informs conferences. I also serve in the editorial boards of a
couple of journals.