I analyze 2D and 3D biomedical images to inform the development of novel drug compounds using deep learning and statistical machine learning. In this role, I created a image-based assay to quantify bone mineral density in trabecular bone, revealing drug response months earlier than existing assays. I enhanced and characterized images of lipid nanoparticles (LNPs) in Transmission Electron Micrographs, leading to a 'go' decision to continue LNP development as a potential siRNA delivery vehicle. I built an automated segmentation application that expedites and largely automates a historically manual task well enough to undergo formal GLP certification toward becoming a new Merck standard practice to quantify drug toxicity. Recently I've used deep learning to segment 3D images to quantify biological 'radiomic' changes from clinical immuno-oncology therapy via CT images, as well as instance segment particles in noisy 2D cryo-EM images to measure and offer views into chemical process re-formulations.

At Rice's Computer and Information Technology Institute, I assisted HPC computing users throughout the university in their use of parallel systems, especially the Rice Terascale Cluster (RTC). I composed and delivered a series of tutorials on High Performance Computing to entice Rice researchers to use the RTC. I also redesigned and built the RTC's web page.

At the Center for Advanced Computing (CAC), as an NPACI Computational Scientist, I enabled scientific computing users of supercomputers such as NPACI’s IBM SP, Cray T3E, and T90, as well as Michigan’s SP2, Origin, and our three Linux Clusters. I also composed and presented numerous HPC software tutorials. I redesigned and built two versions of the CAC’s web site, composing all of the written content.

I consulted at SoarTech to enrich awareness of rising software initiatives in the DoD and write proposals.

As lead engineer at MRJ Technology Solutions (now a division of GDAIS) I was in charge of system infrastructure on a supercomputer-fielded Big Data system (about 500 TB in 2019-equivalent units). There I led the acquisition of our $3M next-generation system. Thus I conceived, developed, and executed a suite of TPC-D-like benchmarks to evaluate data-parallel computers (Cray CS6400, Convex Exemplar, IBM SP2, Sun2000, Connection Machine CM-2, CM-5, and SGI Origin). I assessed the results and chose the replacement of our CM-2: a large SGI Origin with a very high throughput fibre-channel RAID 3. This led to our acquisition of a $3 million ultra-high performance next generation Big Data platform, and the subsequent porting of our custom data-parallel column-based database application thereto. (In brief, it was a data-parallel task scatter-gather system not unlike MapReduce and Hadoop, but a decade before.)I also designed and built a high-performance database I/O library that moves data in parallel to/from our RAID-based database, asynchronously overlapping disk access with computation (via POSIX threads on a 32-CPU Silicon Graphics Origin 2000).I wrote and performance-tuned numerous parallel applications in support of a VLDB fielded in C* on a CM-2 Connection Machine and written in C with POSIX threads on a SGI Origin, and I participated in the porting of this large custom-implemented data warehouse and data-mining system from our CM-2 to the Origin. I was also responsible for supporting an academic C* compiler and communications library that underpinned the database and was based on POSIX threads, which replicated the CM-2’s dataflow services onto the SGI Origin.I assessed cutting-edge technologies and products (from data storage media to processor architectures and interconnects to gigabit networks) for their potential usefulness in serving the future needs of the project. I also wrote or edited most of the project’s documents.

In the Artificial Intelligence Technical Center, I designed and developed the 2D graphical tactical overlay server that was part of the US Army’s Terrain Evaluation Module (TEM) GIS system. I also designed and built TEM’s network socket layer that allowed the system to act as client, server, or both, in concert with its X services, in order to ideally integrate TEM with existing and future Army external tactical decision support applications. I also administered a network of HP/UX workstations.In MITRE’s Air Traffic Control Federally Funded R&D Center for the Federal Aviation Administration, I supported research into flight delay propagation throughout the North American airspace. And I installed and administered a LAN of Sun workstations and a multiprocessor Solbourne (Sun clone) server.