Austin Research Laboratory, Austin, TX, USA
I am currently involved in several projects that seek to apply HPC concepts and systems to other domains. One project I am working on is a collaboration with others at the Austin Research Lab and MD Anderson Cancer Center to reduce the time from tumor identification to treatment for proton therapy patients from days to mere minutes utilizing new algorithms and parallel computing platforms (http://www.research.ibm.com/articles/proton-radiation-therapy.shtml). Another project funded by the Department of Energy is looking at ways to provide online Performance Health Monitoring for large-scale systems such as the PERCS or BG/Q systems from IBM. Finally, I am involved with the Genomics Strategic Initiative that seeks to research novel solutions to enable the effective use of data from human genome sequencing, transitional medicine, and personal health care solutions. From 2/2011 to 2/2012 I served as manager of the Workload Optimized System group here in the Austin Research Lab. My group focused on the development of architectural features and system software that allow inexpensive customization for each potential customer's needs. Specifically, we improved the performance of "big data" business analytics software packages through performance modeling, analysis, and system design. Our innovations on solutions to support packages such as Big Insights allow customers to analyze petabytes of semi-structured or unstructured data quickly and accurately. From 2003 to 2011 I worked on the PERCS project here at IBM, developing a POWER 7-based supercomputer under the HPCS program sponsored by DARPA. I researched key areas of computer architecture that appear in the POWER7 processor, distributed systems as embodied in the high-end P775 offering from IBM, and modeling the performance of such large-scale systems. I have 33 patents relating to my work on PERCS and other projects here at IBM, and have published 30 papers, including publications in top conferences such as ISCA, HPCA, and ICS.