Computational Research Scientist
Current- DASSA, a scalable data analysis engine and signal processing library (like Matlab Signal Processing toolbox) for terabyte data from Distributed Acoustic Sensing (DAS) https://bitbucket.org/dbin_sdm/dassa/src/master/ (/w permission)• FasTensor, a structure locality-aware parallel data processing system towards replacing MapReduce/Spark paradigms for multi-dimensional array processing, such as convolution used by CNN. It is 2000X faster than Spark and 26%X faster than TensorFlow. https://bitbucket.org/berkeleylab/fastensor/src/master/.• SDS, an autonomous data management framework for optimally selecting data index/organization to accelerate terabyte or even exascale-byte data analysis and mining on supercomputer. https://code.lbl.gov/svn/sds/ (/w permission)• DataElevator, an ad-hoc software to perform low-contention data movement in deep-hierarchical storage system. https://bitbucket.org/sbyna/dataelevator.• ExaHDF5, a HDF5 array data model based library to scale its key features to upcoming exascale super computing. https://sdm.lbl.gov/exahdf5/• Proactive Data Containers (PDC), an object-based data abstraction for emerging HPC systems which are expected to be deployed with an unprecedented level of complexity, due to a very deep system memory/storage hierarchy. https://sdm.lbl.gov/pdc/• SDS-Sort, the first parallel sorting algorithm with upper bound of O(4Np ) on load balancing, scaling to 130K CPU cores and having sorting throughput of 117 TB/min.
