Postdoc in High Performance and Scientific Computing
We are seeking a well-motivated postdoc in one or both following areas of High Performance and Scientific Computing. The responsibility of the candidate will be to contribute research in at least one of the following:
Autotuning and automatic optimization methods for CPUs and accelerators (primary GPUs, we are also interested in MIC or FPGA). We are currently focusing to online autotuning for heterogeneous computing nodes. The main challenge here is the development of a fast tuning space search method, which can tune program online with partial knowledge of program behavior with different setup (e.g., different hardware or input). We are also developing automatic source-to-source transformation methods allowing fusion of CUDA or OpenCL kernels and plan to connect autotuning with task-based runtime systems.
Application of autotuning frameworks, task-based systems and source-to-source transformation methods for parallelization, optimization, and acceleration of scientific software from various fields. Seeking a novel approach (that the applicant may bring with him/her e.g., as the result of his/her Ph.D. work) for re-formulation and parallelization/acceleration of scientific algorithms. We can offer a strong interdisciplinary background and collaborations in the area of computational chemistry (molecular docking and molecular dynamics) and cryo-electron microscopy. We further plan to connect autotuning with deep-NN frameworks.
Job characteristics and specifics:
Perform research in the area of HPC and Scientific Computing
Publish in high-quality journals and conferences
Support Ph.D. and undergraduate students
Identify and pursue new research lines
You should have:
Ph.D. degree in Computer Science or Mathematics or closely related field
Publication record in the area of our interests
Excellent autonomous thinking and problem-solving skills