Professional Experience

Research Scientist, Rice University, Houston, TX, USA    2015-present

Responsible for conducting and leading research on 1) LLVM-based optimizations for large-scale systems, 2) code generation and optimizations for Graphics Processing Units (GPUs), and 3) automatic construction of compiler/runtime heuristics using machine-learning techniques.

CAS Research Fellow, IBM Centers for Advanced studies, TORONTO, CANADA    2014-present

Responsible for leading an IBM CAS research fellowship project titled “GPU Enablement of Java Applications”, which aims to add a capability of generating GPU code to IBM’s Java compiler and virtual machine. I personally work on 1) compiling and optimizing Java 8 programs for GPU execution, and 2) constructing performance heuristics that select a preferable computing resource using machine-learning techniques.

Postdoctoral Researcher, Rice University, Houston, TX USA    2013-2015

Conducted research in the area of compiler and runtime support for parallel languages such as Habanero-Java. In particular, I worked on high-level language constructs and parallel code generation for GPUs while maintaining precise exception semantics on GPUs.

Assistant professor, Waseda University, Tokyo, Japan    2012-2013

Conducted research on automatic parallelizing compiler and compiler-directed power reduction techniques in collaboration with many companies in Japan.  More specifically , I worked on parallelizing and optimizing 1) automotive engine control softwares with TOYOTA and Denso,  2) a dose calculation software for heavy-ion therapy with Mitsubishi Electric, 3) smartphone applications on Android platform with Fujitsu Ltd., and 4) multimedia applications with Olympus Corporation.

Instructor for freshman-level 1) computer literacy class and 2) programming class with C language in Spring/Fall 2012.

Research Associate, Waseda university, Tokyo, Japan — 2010-2012

Conducted research on automatic parallelization and power reduction techniques for embedded heterogeneous multi-core processors. In addition, I worked on the automatic parallelization and optimization of a clinically-used dose calculation software for heavy-ion therapy in collaboration with Mitsubishi Electric. Based on experimental results on an IBM POWER7 platform, our optimized version was over 50x faster than the conventional version. 

Research Associate , Global COE PROGRAM, Waseda University, Tokyo, Japan — 2007-2010

Conducted research on automatic parallelization and power reduction techniques for embedded heterogeneous multi-core processors in collaboration with Hitachi Ltd. and Renesas Electronics Corporation (formerly known as Renesas Technology). This work was partially supported by one of the global centers of excellence (GCOE)  program named  “International research and education center for Ambient SoC” established by Japan’s Ministry of Education, Culture, Sports, Science and Technology (MEXT).