Ben Huntington Email and Phone Number

Developing AI and Physics simulation solutions for leak detection on pipelines. - Researched, developed, and implemented machine learning based solution for leak detection in shut-in pipelines - Developed and implemented signal analysis solution for finding leak location on pipelines.-Researched, developed, implemented, and tested methodology for using flow simulation to assist in leak detection on poorly instrumented pipelines- Created internal tools for tuning two different leak detection softwares, these tools use a combination of data visualization, machine learning, and statistical analysis to speed up, simplify, and standardize deployment of our software- Supporting deployment of our leak detection software-Training new engineers how to deploy our software on clients' pipeline systems

Studying and analyzing data for projects that are of interest to me.Accomplishments• Completed multiple online data science courses• Performed end to end data analysis on Kaggle data by doing data wrangling, feature engineering, and data visualization to analyze data and then make predictions by training and testing machine learning models such as neural networks and RandomForest using Python.Project: Time Series Classification Competition • Wrote Python code to do feature engineering and label encoding to prepare time series collected from a robot driving on different surfaces. This data has 10 features and time series have lengths of 128 data points.(See https://github.com/BHuntington7/data-projects/tree/master/Kaggle%20Robot%20Competition)• Trained random forest models for classification on features produced by tsfresh, a time series analysis Python library• Ranked 156 out of 1478 competitors Project: House Price Prediction • Wrote Python code to do data cleaning, feature engineering, and label encoding to prepare a house pricing data set for processing (1460 samples in 80 features). (See https://github.com/BHuntington7/data-projects/tree/master/House%20Price%20Analysis)• Trained and evaluated multiple random forest models based on a random grid of hyperparameters and obtained optimal hyperparameters • Used the model to predict prices and obtained an overall root mean squared error on the test set is 0.14359.Project: Natural Language Processing (NLP) for Finding Equivalent Questions• Analyzed similar and dissimilar questions from a set of 168,000 questions from the AskUbuntu data set• Pre-processed raw text to enable batch processing in Pytorch by using padding and pre-trained word embeddings• Built and trained a CNN model with an embedding layer, a 2D convolution layer and a 1D max pooling layer to map each question to a numerical vector and then used a cosine similarity function to classify the question pairs as similar or dissimilar

Analysis, Testing, and Algorithm Development for a financial AI product.

Mathematical modeling, coding, and data analysis work.Accomplishments• Used multivariable calculus (Lagrange multipliers in 100s of dimensions) to optimize a weighted averaging scheme and implemented this scheme in a large C++ code and a large Fortran code using Peridynamics. • Developed a series of tests for the new capability using gtest• Implemented parallel algorithm to speed upcomputation using MPI in Fortran.• Profiling and doing performance analysis for a finite element algorithm in C++ using a GPU implementation.• Generated random porosity materials in Matlab and tested them vs structured porosity materials in simulated mechanical testing.

Worked in Dr. Bonnecaze’s lab doing simulation of multiphase flow in a polymer co-extruder.AccomplishmentsDevelopment of Numerical Simulation of Multi-Phase Flow to Capture Large Interface Deformation• Coded a 3D multi-fluid flow simulator in C++ that uses the Smoothed Particle Hydrodynamics (SPH) method.• Implemented particle search using nearest neighbor search algorithms and solved resulting linear systems using Petsc libraries• Validated code against analytical solution for 3D flow of two fluids in a square channelPolymer Coextruder Feedblock Simulation• Simulated multi-layer polymer flow in a feedblock with up to 8 layers using ANSYS POLYFLOW software• Determined primary cause of polymer interface deformation• Innovated a process improvement that resulted in a superior experimental product and a provisional patentPolymer Coextrusion Die Design• Successfully simulated multi-layer polymer flow in many different die designs with up to 8 layers• Implemented various design principles to reduce interface deformation in simulations• Guided the experimental design process through the use of simulation resultsAnalysis of Roll-to-Roll Coating of Soft Particle Pastes• Formulated an analytical model of roll-to-roll soft particle paste coating• Performed a stability analysis on the process• Determined critical parameters for stability in the process

Modeled and analyzed data from gas liquid separatorsAccomplishments• Modeled and analyzed data to assess the performance of internals in gas liquid separators• Compared vendors’ internals using these models to aid in making business decisions• Updated ExxonMobil’s currently used design models for internals to be more specific and accurate

Worked on process improvement for zirconium (Zr) reduction process. Took on projects to increase yield and quality.Accomplishments• Identified a source of iron contamination to our product Zr, and alterations showed a 2.5% decrease in iron. • Designed, built, and tested a device that uses argon to flush contaminants out of the reactants• Analyzed and recommended modification to thermal processing of Zr product to minimize waste

Led a collaborative project to improve mixing of an organic solvent and water in a 384 well plate using vertical and lateral shaking as a means of mixing.Accomplishments• Created models based on the physics of shaking and its effect on the liquid within the wells• Designed experiments to test these models for improved mixing • Recommended changes to the tool to optimize mixing

Worked on the wafer polishing process. Projects were focused on increased wafer output and quality improvement.Accomplishments• Characterized wafer surface roughness as a function of the polishing process. Experiments showed that it may be possible to remove a step in the process and achieve the same quality.• Linked scratches on wafers to dirty machines. Wrote and enforced a cleaning regimen for operations that reduced the number of wafers rejected for scratches by 2%.• Tested and analyzed a new process path that cleans wafers more thoroughly to prevent contaminants from binding permanently to the wafer surface. This significantly decreased the number of defects and improved quality.