Curriculum Vita

Profile

Dr. J. Ben Deaton is senior data scientist, machine learning engineer, and researcher with 17 years of experience developing and applying mathematical models to solve computationally difficult and meaningful technical problems.

Experience

Shiplify, Atlanta, GA

Chief Data Scientist: Sept 2019–Present

• Lead machine learning work in transportation/logistics startup
• Primarily focus on application of computer vision (CV) and natural language processing (NLP) deep learning models to various problems in the shipping/logistics space

Coherent Path, Boston, MA

Data Scientist: July 2016–Sept 2019

• Worked directly with mathematician CTO to develop and productize core machine learning and statistical algorithms
• Broad experience ranging from exploratory data analysis/visualization, data cleaning, feature engineering, experiment design, reporting, model development/validation, and writing production machine learning code

Simpson, Gumpertz & Heger, Boston, MA

Consultant, March 2013–July 2016

• State-of-the-art consulting using computational models (nonlinear finite element analysis, neural networks, etc.) to predict crack propagation and failure modes of critical reinforced concrete infrastructure subjected to seismic, wind, and man-made (blast and missile impact) effects
• Computational modeling and risk assessment of U.S. Department of Energy (DOE) and commercial nuclear facilities
• Neural network analysis of wind-induced failure of U.S. Department of Defense (DOD) structures
• Automated data processing and analysis, generation of models, and post-processing results of large-scale simulations

Georgia Tech College of Engineering, Atlanta, GA

Lecturer, 2009–2012 (during grad school)

• Instructor of record for statics course with full course responsibility
• Eight (8) semesters experience
• Cumulative course survey rating of 4.73/5.0

Georgia Tech Computer Aided Structural Engineering (CASE) Center, Atlanta, GA

Software Developer, 2003–2006 (Part-time)

• GT STRUDL (Commercial Finite Element Software) features developed:
  • DESIGN SLAB command to design RC slabs based on FEA
  • GTMESH module to generate mesh for arbitrary domains using Delauney triangulation and custom quadrilateral mesh algorithm

Education

Georgia Institute of Technology, Atlanta, Georgia

Ph.D, 2012, Structural Mechanics

• Dissertation: Nonlinear Finite Element Analysis of Reinforced Concrete Exterior Beam-Column Joints with Nonseismic Detailing
• Minors completed: Solid Mechanics, Mathematics
• Advisors: Drs. Kenneth M. Will and Lawrence F. Kahn

M.S., 2005, Civil and Environmental Engineering

• Thesis: A Finite Element Approach to Reinforced Concrete Slab Design
• Advisors: Drs. Kenneth M. Will and Lawrence F. Kahn

B.S., 2003, Civil and Environmental Engineering

• With Honor

Courses and Certificates

Coursera: Machine Learning (Stanford/Andrew Ng)

Credential ID YLXNXTVKF9ZL, Issued February 2020.

Publications

1. Contributor, “ACI 447R-18: Design Guide for Twisting Moments in Slabs”, ASCE/ACI Joint Committee 447 (Finite Element Analysis of Reinforced Concrete), American Concrete Institute, 2018. [www]
2. Deaton, J.B., A.T. Sarawit, and S. Bolourchi, “Three-dimensional Nonlinear Modeling off Reinforced Concrete Slab-column Connections with Unreinforced Column Capitals,” 24rd International Conference of Structural Mechanics in Reactor Technology (SMiRT-24), Busan, Korea, August 2017.
3. Pridmore, A., Bian, S., Engindeniz, M., Deaton, J.B., and R.P. Ojdrovic, “DC Water 22-Feet Brick Sewer Emergency Repair,” ASCE Pipelines Conference, Kansas City, MO, 2016.
4. Deaton, J.B., A.T. Sarawit, and S. Bolourchi, “Nonlinear Force-Drift Backbone Curve for Reinforced Concrete Squat Shear Walls,” 23rd International Conference of Structural Mechanics in Reactor Technology (SMiRT-23), Manchester, UK, August 2015.
5. Deaton, J.B., “Nonlinear Finite Element Analysis of Reinforced Concrete Exterior Beam-Column Joints with Nonseismic Detailing,” Ph.D. Thesis, Georgia Institute of Technology, 2012.
6. Deaton, J.B. and K.M. Will, “Nonlinear FEA of Cyclically-Loaded Exterior RC Beam-Column Joints with Inadequate Seismic Detailing,” 15th World Conference on Earthquake Engineering (15WCEE), Lisbon, Portugal, Oct. 2012.
7. Deaton, J.B. and L.F. Kahn, “Identification of Reinforced Concrete Failure Modes using Linear Elastic Finite Element Analysis,” 6th Congress on Forensic Engineering, ASCE, San Francisco, CA, Nov. 2012.
8. Shin, M., A. Bommer, J.B. Deaton, and B. Alemdar, “Twisting Moments in Two-Way Slabs: Design Methods for Torsion in Slabs using Finite Element Analysis,” Concrete International, Vol. 31, No. 7, July 2009, pp. 43-48.
9. Deaton, J.B., “A Finite Element Approach to Reinforced Concrete Slab Design,” Master’s Thesis, Georgia Institute of Technology, 2005.

Presentations

1. Deaton, J.B. and K.M. Will, “Calibration of Material Models for NLFEA of Seismically Deficient RC Beam-Column Joints,” American Concrete Institute Spring Convention, Milwaukee, WI, 2016.
2. Deaton, J.B., “Nonlinear Finite Element Analysis of 3D Reinforced Concrete Beam-Column Joints,” TNO DIANA, 2013.
3. Deaton, J.B., “Nonlinear Finite Element Analysis of Exterior Corner Reinforced Concrete Beam-Column Joints,” Engineering Laboratory, National Institute of Standards and Technology (NIST), Gaithersburg, MD, 2013.
4. Deaton, J.B., “Nonlinear Finite Element Analysis of Reinforced Concrete Beam-Column Joints,” Geotechnical and Structures Laboratory (GSL), US Army Corps of Engineers (USACE) Engineer Research and Development Center (ERDC), Vicksburg, MS, 2012.
5. Deaton, J.B. and L.F. Kahn, “Lessons Learned from Forensic FEA of Failed RC Structures,” talk, American Concrete Institute Fall Convention, Pittsburgh, PA, Nov. 2010.
6. Deaton, J.B., K.M. Will, and L.F. Kahn, “Accuracy of Finite Element Based Slab Design Moments,” talk, American Concrete Institute Fall Convention, St. Louis, MO, Nov. 2008.
7. Deaton, J.B. and K.M. Will, “Mesh Generation for Arbitrary Domains in GT STRUDL,” 20th Annual GT STRUDL User’s Group Meeting, Las Vegas, NV, June 2008.
8. Deaton, J.B. and K.M Will, “A Finite Element Approach to Reinforced Concrete Slab Design in GT STRUDL,” 18th Annual GT STRUDL User’s Group Meeting, Atlanta, GA, June 2006.

Honors

• Georgia Tech CETL/BP Outstanding Graduate Student Instructor Award, 2011
• Georgia Tech CEE Bill Schutz Graduate Teaching Assistant Award, 2011

Service

Coming soon.

Software Proficiency

Data Science / Software Engineering

• Programming languages: Python, Fortran
• Machine learning:
  • Traditional: scikit-learn, Statsmodels, Matlab Neural Network Toolbox
  • Deep learning: PyTorch, PyTorch-Lightning, Keras, TensorFlow
  • Other: Snorkel (weakly supervised training label generation)
• Distributed systems: PySpark (RDD/SQL/DataFrames)
• Exploratory data analysis and Pipelines: Pandas, Jupyter notebooks, Numpy, Scipy
• Visualization: Seaborn, Matplotlib, Plotly/Dash
• Version control: git, Mercurial
• Software engineering: VS Code, Sublime Text, vim, static analysis/code quality tools (black, flake8, mypy, pytest)
• Mathematics: Matlab, MathCad, Maple, Excel
• Communication: LaTeX, Sphinx, Jekyll
• Cloud: AWS (EC2, S3, SageMaker)

Finite Element Analysis

• Abaqus: Expert user. Experience using CAE for advanced model creation, conducting both implicit and explicit nonlinear analysis, expert user of Concrete Damaged Plasticity and Smeared Crack models. Experience writing user material subroutines and scripting both within Abaqus as well as external scripting for model generation and post-processing.
• LS-DYNA: Experience performing explicit analysis of progressive collapse, impact analysis, and blast loading of structures, use of Winfrith and FHWA concrete models, wrote user material subroutines.
• DIANA: Primary software used during Ph.D. for concrete failure simulation. Expert user of concrete smeared rotating crack model and multidirectional fixed crack models.
• GT STRUDL: Was a software developer for GT STRUDL during grad school, used extensively as consultant for linear analysis and design of structures.
• Ansys: Primary experience using ADPL for seismic evaluation of nuclear facilities and post-processing Ansys output using Python/Pandas.
• FEMAP: Extensive experience building finite element models in this environment for export to and translation between Abaqus, Ansys, LS-DYNA, and DIANA.