Personal Website: http://www.math.wm.edu/~mdbaker0

Student Trainee in Mathematics and Statistics for the National Aeronautics and Science Administration (NASA) in the Climate Science Branch of the Earth Science Directorate, (01/2012 - 12/2012). Part of the Data Management Team for the Cloud and Earth Radiant Energy System (CERES) project.

Member of the CERES Flight Model 6 steering group and CERES processing oversight board. Designed and analyzed changes to the CERES data processing flow, needs of the users, toolkits and libraries, and software configuration change requests. The CERES software project was recently (July 2012) assessed at a CMMI maturity level of 2 and a capability level 3 in all production areas.

Profiled the CERES spatial gridding subsystem and collected runtime statistics between two suborbital satellites Terra and Aqua. Analyzed the benefit of combining main and post processing routines to decrease latency. Scripts written in PERL and csh, code written in Fortran, and statistics and visualization in R.

Optimizing ground data processing cloud retrieval algorithm for multiprocessing at the Atmospheric Sciences Data Center. Initially implemented symmetric multithreading using OpenMP on a highly-tuned XLF compiler, but discovered that POWER6 hardware constraints prevented large gains. Currently using MPICH2 implementation of the MPI-2 standard to distribute 48 hourly jobs across 144 cores with theoretical speedup of x2.7.

Assisting the development of CERES AuTomAted Job Loading SYSTem (CATALYST), scheduled to go live in August, 2012 (Click for process design diagram). In the past CERES data requests were funneled from the PR Web Interface to a manual operator who would pass the raw data records (RDRs) through the process chain resulting in one of CERES' four data products. System parses XML, database queries in PostgreSQL, DRMAA job manager written in Java, subsystem scripts written in PERL.

Proposed using distributed high performance computing on CubeSats to assist in the heliophysics space mission. Demonstrated how multi-core fault mitigation methods using Arduino microprocessors could satisfy the fault tolerance and spatial / technical constraints outlined in the technical readiness levels of decadal survey (cube dimensions, power consumption, etc). Also showed how the cores could be moved on and off line and continue computation without data loss, load managment and scheduling for this cluster critical differentiation point. Discussed the financial impact of using modified COTS hardware versus developing application-specific processors. LED demonstration of internode communication networks.

Future prospective project: Inspired by IBM's Watson, Langley Information Management Branch wants to develop advanced decision support systems for NASA using deep analytics. Space flight resolves issues using Failure Isolation, Discovery, and Recovery (FDIR) and NASA aims to use IBM's Deep Question and Answer (DeepQA) to create a highly sophisticated failure isolation system by structuring and reasoning over NASA's large body of natural language content, then generating ranked answers with associated confidences.

(edit description)