Abstract:

Rapidly growing importance of computational and data-centric methods in science and engineering, as well as the drastic transformations in computer architecture open up exciting opportunities for research in parallel computing. I am mainly interested in the design and development of algorithms and software systems that can harness the full potential of state-of-the-art computing platforms to address challenging problems in data-intensive scientific computing.

I will start my talk by describing a highly scalable iterative eigensolver that we developed for large-scale sparse matrices. I will present performance improvements obtained by the use of our solver for nuclear structure computations. Sparse matrix computations and eigenvalue problems arise in various fields of contexts, such as graph analytics and machine learning among others, but people working in these fields are not necessarily parallel programming experts. So in the second part of my talk, I will describe our work on the development of a software ecosystem, called DOoC+LAF, to enable high performance and productive graph and matrix analytics on emerging non-volatile memory architectures. Last but not the least, I will talk about my recent efforts towards creating a fully automated force field generation and optimization framework in collaboration with the Materials Project group at LBNL. The goal here is to leverage the extensive computational and experimental data that is becoming available under the Materials Genome Initiative for enabling highly accurate molecular modeling and simulation. By developing new techniques in materials informatics, this ambitious project has the potential to change how computational scientists design advanced materials and nanotechnology products.

Bio:

Hasan Metin Aktulga, a native of Turkey, received his B.S. degree from Bilkent University (2004), M.S. and Ph.D. degrees from Purdue University (2009 and 2010), all in Computer Science. He is currently a postdoctoral research fellow with the Scientific Computing Group at the Lawrence Berkeley National Laboratory.

Dr. Aktulga's research focus is in the area of parallel computing, with emphases on high performance computing, data-intensive computing and scientific computing. He is interested in the design and development of parallel algorithms, numerical methods and software systems that can harness the full potential of state-of-the-art computing platforms to address challenging problems in large scale data-intensive scientific applications. Dr. Aktulga's work has been published in prestigious conferences and journals. His Ph.D. work on parallel reactive molecular dynamic simulations has been recognized by ScienceDirect among the Top 10 articles published in the Parallel Computing Journal in 2012 and among Top 25 in 2013, and has consistently been among the 10 most downloaded articles in Nanoscience and Nanotechnology Commons since 2012. His conference publications have been nominated for best paper awards in HPCS 2011 and SC 2013 conferences. Dr. Aktulga has served as a program committee member and reviewer for several conferences.