The ol' career path debacle. Any insight helpful

[chusteczka]

OP, you are not an elitist, your current university does not matter to this discussion, and your focus is clearly on mathematical computational theory rather than production. There is nothing wrong with your direction, in fact it is excellent. I completely understand the direction you are going in even if it is above my mathematical skills.

At my university, we discussed three American universities worthy for graduate study:

1. MIT
2. Berkeley
3. University of Illinois at Urbana/Champaign

Here are the lists for the top 5 computer science graduate programs.

• MIT
• Stanford
• Berkeley
• Carnegie Mellon University
• University of Illinois at Urbana/Champaign

http://pages.cs.wisc.edu/~remzi/rank.html
http://grad-schools.usnews.rankings...schools/top-computer-science-schools/rankings
http://grad-schools.usnews.rankings.../top-engineering-schools/computer-engineering


Any of those top five should be a good fit for you.

It appears to me that you need more classes in statistics, machine learning, numerical methods, and parallel computation. Probably even more on algorithmic theory.

I know for a fact that the University of Illinois at Urbana/Champaign (UIUC) has what you need.

Here is the CS course catalog at UIUC.

The numerical methods class is taught by Professor Michael Heath. Here are the classes Professor Heath teaches. He wrote this book; Scientific Computing: An Introductory Survey. Professor Heath has a top reputation as a professor and graduate adviser.

Professor Gropp as featured on the UIUC department's main page is a leading researcher in parallelization. He developed the Message Passing Interface (MPI) at Argonne National Laboratory and has only recently transferred from Argonne to UIUC. There is much collaboration between UIUC and Argonne National Laboratory that has resulted in large-scale parallel computer purchases for UIUC.

My advice is to apply at all five of those schools plus a couple others as a backup since it is extremely difficult to get into any of those schools at the graduate level. For instance, UIUC accepts only the best students from around the world.

My University is world-class, I daresay top 5 in the world for Machine Learning/Statistical Learning.

I question the accuracy of the above statement if there are only 20 students in the graduating class. Maybe because my graduating class for computer science had several hundred and can certainly be figured out by the information provided above. I have a feeling you did not get too in depth in any of those topics but you do seem to have a good conversational handle on them.

Another problem is that I am so in love with Operating Systems, Processor Design (all the low level GUTS) that I am having a hard time choosing what to do. I really love performance modelling and learning about why certain parellelization models failed and succeeded.

Do not worry about focusing in a specific area to the detriment of other classes. Those other classes are all basic and remain useful no matter which theoretical direction you take.


Edit:
Additionally, UIUC has a strong distance learning program where classes can be taken over the internet. One of my classes had a Colonel from the Pakistan army as a fellow student. It was pretty funny when he introduced himself on the discussion forum.


Edit 2:
I missed this part.

I am really torn between the idea of highperformance systems/Virtual Machine research on Shared Memory Systems/Processor Design/Operating Systems type research, or Statistical Learning/Machine Learning.

The Low Level Virtual Machine (LLVM) project is based at UIUC.