Tom Stephens Email & Phone Number

Arlington, Virginia, United States

Arlington, Virginia, United States

Scientific, Engineering, and Technical Assistant to DARPA's Information Innovation Office and Tactical Technology Office

Durham, North Carolina, United States

Developed, Led, and Executed 6.1, 6.2, and 6.3 level research programs in machine learning for US Govt agencies.

Durham, North Carolina

Bethesda, Maryland

Independent Contractor (comparable to post-doc)A surprisingly large number of biological functions at the cellular and subcellular levels are facilitated by the deformation of the soft, flexible membranes that encase our cells and their organelles. In healthy cells, membrane-wrapped packages holding chemical signaling molecules are routinely delivered to the cell boundary and scheduled for release into the outside world. As well, viruses on the exterior of a cell coax the plasma membrane into engulfing them, defeating the cell's primary passive defense and advancing their mission of replication. These membrane-mediated processes are excellent pathways for drug delivery and disease prevention, making working models of dynamic membrane remodeling immediately applicable. My work at NIH involved:Modeling biophysical drivers of membrane deformation using continuum models and simulationsDesigning and implementing finite element methods for deforming surfacesWorking on analysis and interpretation of dynamic geometric membrane features in the context of biological function

Bethesda, Md

An important strategy in biomedical research is the selective alteration of individual proteins, which furthers mechanistic understanding of their role and ultimately permits the development of targeted therapies. Testable, repeatable protein alteration requires knowledge of the spatial relationship between the amino acids that make up the particular protein. Sequences of amino acids along a protein backbone can be deduced through painstaking analysis of their nuclear magnetic resonance (NMR) spectrum. I was introduced to this area of structural biophysics during the summer of 2014 and:Combined FFT cross-correlation and k-nearest neighbors strategies to develop an iterative image registration algorithm for the alignment of two-dimensional projections in four-dimensional NMR data setsIncorporated a regular expression (regex) procedure to assist in assigning resonance data peaks to known amino acid sequences in several small proteins Collaborated with NMR spectroscopists throughout the development cycle in order to obtain a practical, working algorithm for real data setsPublished algorithm and comparison to existing methods in the Journal of Biomolecular NMR

Fairfax, Virginia

Bounded, invariant sets in the phase space of a system of ordinary differential equations correspond to fixed points, periodic and heteroclinic orbits, and otherwise organized phenomena. In textbook examples, the initial conditions or model parameters that yield such predictable behavior can be discovered through more or less sophisticated analytical tricks. In the real world, these features may only be discoverable using computational tools. My thesis work combined abstract results from algebraic topology with easily computable information to identify invariant sets for a wide class of ordinary differential equations. By the end of the project, I had:Designed and implemented adaptive subdivision algorithms for domain decomposition, yielding cubical sets representing superlevel and sublevel sets of continuous, real-valued functions on domains of arbitrary finite dimensionFormulated and applied vector field segmentation criterion permitting the construction of smooth manifolds satisfying first and second fundamental form constraints (with respect to the ambient vector field) ___________________Nucleation and spinodal decomposition are two mechanisms of phase separation in metal alloys. These processes work to decompose a well-mixed material into regions consisting entirely of one metal component or another. Watching this process unfold reveals a changing pattern made by the distribution of the individual metal components in the alloy (evolving from a fine-grained structure to often fantastically complex macroscopic, interlocking features). These evolving patterns are a perfect setting in which to develop computational (and quantitative) descriptions of shape. During my time on this project I:Designed metrics to obtain quantitative distinctions between between large data sets of time-evolving patterns using tools from computational topology Implemented graph algorithms to count connected components and cycles for two-dimensional binary image data

Phd Mathematics, Algebraic Topology, Dynamical Systems, Partial Differential Equations

Activities and Societies: Mentoring undergraduate researchers, delivering summer workshops, teaching classes, and installing linux on old.

Master’S Degree, Mathematics

Bachelor’S Degree, Mathematics, Minor In Computer Science

Activities and Societies: Undergraduate Research, Tutoring, Grading