Highlights:

Skills:

Ph.D. Researcher 09/2014 - current
University of Notre Dame, , United States
Participated in multiple computational material science research projects. Featured in 6 peer-reviewed scientific journal publications (5 as lead author) and presented at 7 international conferences.
Selected projects: - Utilized Hybrid Monte Carlo-Molecular Dynamics Methodology to simulate environment-dependent behavior of weakly charging polymers, including complexation properties. - Developing atomistic simulation using Artificial Neural Networks (ANN) for additive to polymer films - Effect of plasticizers on conformation and H-bonding. - Served as a developer to implement algorithms for Monte Carlo simulation in SAPHRON (a C++11 Monte Carlo library) and connected it to open-source Molecular Dynamics package (LAMMPS) to enable faster and efficient sampling of states of complex molecules. - Implemented state-of-the-art advanced sampling algorithms in C++ software package SSAGES for free energy calculations in molecular simulations - predicting performance of water filtration membranes. Machine Learning, Data Science and AI Projects: Which dog breed do you look like? | Convolutional Neural Network & Transfer Learning | Link: Built and utilized Convolutional Neural Network (CNN) architectures to estimate canine’s breed. If human image was supplied, the resembling dog breed was identified. CNN training accomplished using Keras with Tensorflow backend with 8351 dog images. Transfer learning with pre-trained CNN architecture (Resnet-50) was also utilized. OpenCV was utilized for detecting human faces. Classification accuracy ~82% was achieved. How much taxi fare will it be in NYC? | Kaggle Competition & Deep Neural Network | Link: Predicted taxi fares in the New York city utilizing Deep Neural Network (DNN) regression model (Keras with Tensorflow backend) and 1 million rows of data with 7 features (pickup/dropoff coordinates, time, date etc.) for training (performed on cloud GPU machine). Simple linear model was utilized for benchmark and RMS Error for performance evaluation. DNN model showed improved performance with RMS Error decreased by 40% as compared to the benchmark. Where to open a restaurant in Toronto, CA? | Unsupervised Learning & K-means | Link: Provided recommendations for opening a new restaurant in Toronto, CA. K-means clustering algorithm was utilized to cluster areas in Toronto (104 Postal Codes) with high density of tourist places and cluster the properties in Toronto based on prices (25,000 rows of data). Then, analyzing the crime & income (141 data points) and restaurant competition statistics (Foursquare API), finally, provided the 2 locations that are ideal for opening the new restaurant.--

University of Notre Dame 09/2014 - 09/2019
, , United States
Degree: Doctoral Degree
Major:Chemical and Biomolecular Engineering
Participated in multiple computational science research projects. Featured in 6 peer-reviewed scientific journal publications (5 as lead author) and presented at 7 international conferences.

Login to view resume: Vik Rathee -