I have some general advice questions, but I'll make them as spec

asked 01 Jan '10, 00:00

Miles's gravatar image

Miles
26028
accept rate: 0%

11 Jan '12, 17:34

Paul%20Rubin's gravatar image

Paul Rubin ♦♦
14.6k412

Not really an advice, but thumbs up for not cutting corners and focus on large scale optimization, which can be hard work. I think you will do great with such skills, but my opinion is a little biased though :-) Best of luck.

(08 Jul '10, 08:18) Bo Jensen ♦

The question – in its full length – was:

I have some general advice questions, but I'll make them as specific as I can. I'll soon have a BS in applied math and an MS in stat, and from everything I've seen, OR looks like a great fit for me. I'm interested specifically in large-scale optimization problems, which I've done some work in at a DOE lab. I'm really on the OR/optimization side, not the operations management side. I'm also a programmer, and I have a soft spot for using large amounts of computing power. (And not with Excel.)

So my questions:

  • Are there a lot of opportunities in this field outside of government labs? I'm interested mostly in industry.

  • Are there any specific suggestions for graduate (doctoral) programs that have the slant I'm looking for?

  • And more immediately relevant, is there any chance of getting a job in something related for a few years before I pursue a PhD?

Thanks for your help. -- @Miles

(01 Sep '12, 03:55) fbahr ♦

There are great opportunities in industry. IBM's Center for Business Analytics and Optimization does great stuff. And, of course, places like google and amazon do lots of interesting large scale problems.

As for doctoral work, places like Georgia Tech (and sometimes my own place Carnegie Mellon) have the chance to do practical work, with a large scale optimization emphasis.

link

answered 08 Jul '10, 02:19

Michael%20Trick's gravatar image

Michael Trick ♦♦
4.1k41533
accept rate: 20%

1

Just wanted to point out that Georgia Tech has an extremely hard entrance exam for the PhD student. What I heard is that less than 40% pass the first time.

(08 Jul '10, 04:42) Mark ♦
2

Most programs have some sort of qualification exam (typically taken a couple of years after start of study). Tech's exam is, as far as I know, not harder or more rigorous than that of other programs. Certainly enough people graduate from Tech to suggest lots of people getting through the process.

(08 Jul '10, 10:46) Michael Trick ♦♦

Try Google, it doesn't get larger scale than that. We have an operations research team and, in general, we solve a lot of large scale problems. We have an internship program if you want to see what industry is like before you graduate.

link

answered 09 Jul '10, 14:49

Emilie%20Danna's gravatar image

Emilie Danna
691
accept rate: 0%

To add to Michael's answer, I believe that AT&T, GE, and Exxon Mobil would also be interesting places to work. They certainly have very tough, large-scale optimization problems to solve.

link

answered 08 Jul '10, 14:43

Tallys%20Yunes's gravatar image

Tallys Yunes ♦
2.1k518
accept rate: 11%

Speaking of Exxon Mobil: BP America ;-) For instance, Data Analyst @ simplyhired.com: http://www.simplyhired.com/a/jobs/list/q-data+analyst+company%3A%28BP+America%29

(08 Jul '10, 16:21) fbahr ♦

Seems to me that once upon a time, maybe before AT&T was broken up, their long lines (long distance) division had one of the largest MIP models (or maybe LP; my memory is slipping) in captivity, used to route long distance calls I think -- and they had to solve it repeatedly in real time.

(09 Jul '10, 21:00) Paul Rubin ♦♦

Airlines, shipping companies, and military logistics also have serious large-scale OR problems.

link

answered 08 Jul '10, 16:25

Matthew%20Saltzman's gravatar image

Matthew Salt... ♦
4.7k310
accept rate: 17%

I am clearly biased, but Georgia Tech may be a particularly appropriate program to consider for those seeking a Ph.D. in OR with an interest in a career in industry.

I do not know how many total positions are out there, but I can tell you that we graduate between 20-30 Ph.D.s yearly, and roughly half of them take industry positions. My own students, and students of close colleagues in logistics, have recently taken positions at: BNSF Railroad, SAP, JDA Software, Manhattan Associates, amazon.com, Goldman Sachs, and ExxonMobil.

link

answered 13 Jul '10, 03:46

Alan%20Erera's gravatar image

Alan Erera
1.0k37
accept rate: 12%

Working for over a decade for ILOG and now Gurobi, I've seen my share of large-scale optimization applications. First, note that large-scale is not synonymous with difficult. For example, we are currently working with one client that has one of the largest models I have ever seen; in this case, the primary challenge involves memory, not computational difficulty.

That said, many of the usual suspects have been identified already: logistics, transportation, supply chain, government. I would add finance and energy to this list. Plus there are some interesting large applications in retail and other service industries.

I suggest you attend an INFORMS conference and focus on application sessions (as opposed to theoretical sessions). This will give you some indication of the projects that people are doing in support of industrial applications.

As for schools, you should look for a program that has significant industry research - projects that involve business sponsors. Georgia Tech is one great program, but there are many other fine choices.

link

answered 27 Oct '10, 18:02

Greg%20Glockner's gravatar image

Greg Glockner
43716
accept rate: 20%

I couldn't agree with this more. Large-scale is not synonymous with difficulty -- problem structure is far more predictive of difficulty. In fact, in nonlinear optimization, introducing lots of auxiliary variables (lifting) tends to help accelerate convergence rather than hinder it. Also, I'd rather solve a large-scale convex problem than a small nonconvex problem (with no idea what the initial guess should be) any day.

(28 Oct '10, 11:36) Gilead ♦

I'm currently working on an HPC optimization project, so I am definitely interested in the computational issues.

(28 Oct '10, 23:41) Miles

I just started working towards a Master's in CS/OR so take my advice with a grain of salt.

I've been to a few career fairs and it seems to me as if there are plenty of companies that are interested in hiring a person with a background in large scale optimization. As expected, most of these companies already know the benefits of large scale OR models, such as IBM, RAND and Amazon.

In terms of degrees, there will be a bunch of programs that you should look into (most of the top departments always have at least one or two people working with large scale OR). Georgia Tech is probably a safe bet in this regard, but it is also the largest OR department in the country and it seems to me as if you could easily be "lost" among so many other PhD candidates.

One last piece of advice is to that experts in large scale optimization usually focus on improving the efficiency of their models and algorithms, and somewhat ignore the effects of uncertainty. While this is fine, it also makes your skills slightly less applicable in industry. So if you do eventually get a degree, I would advise you to look into fields such as Stochastic Programming or Dynamic Programming, which will probably broaden your skill set (not to mention that knowing how to solve such problems will also allow you to do research outside of the OR department).

link

answered 28 Oct '10, 00:09

Berk%20Ustun's gravatar image

Berk Ustun
3071311
accept rate: 0%

edited 29 Oct '10, 02:55

Having worked in the airline industry for a long time where we routinely solve those gazillion decision variable scheduling problems, retail (where i currently do OR work) makes that look puny. Retailers collect data like crazy and when you look at their chain-level optimization problems, you being to see the scale. E.g. a chain has a 1000 stores in the US (and a smaller number of warehouses, distribution centers etc), and maintains 10-100,000 SKUs (stock keeping units or unique products). Combine this with the monstrous detail of data available and we have a perfect optimization storm. If that is not enough, customers purchase combinations of SKUs (shopping baskets), and these SKUs can be complements or substitutes, so the true number of decision variables in a particular instance we look at is astronomical.

link

answered 11 Nov '10, 15:35

shiva%204's gravatar image

shiva 4
2635
accept rate: 11%

Your answer
toggle preview

Follow this question

By Email:

Once you sign in you will be able to subscribe for any updates here

By RSS:

Answers

Answers and Comments

Markdown Basics

  • *italic* or _italic_
  • **bold** or __bold__
  • link:[text](http://url.com/ "Title")
  • image?![alt text](/path/img.jpg "Title")
  • numbered list: 1. Foo 2. Bar
  • to add a line break simply add two spaces to where you would like the new line to be.
  • basic HTML tags are also supported

Tags:

×190
×51
×38

Asked: 01 Jan '10, 00:00

Seen: 3,770 times

Last updated: 01 Sep '12, 03:57

OR-Exchange! Your site for questions, answers, and announcements about operations research.