Data Machine

Rohit Bisht

Kearney, NJ-***** *****@****.*** 201-***-****

LinkedIn: https://www.linkedin.com/in/rohit-bisht-a671a482/ OBJECTIVE:

Data scientist willing to learn new technologies with commendable skills in statistical languages, machine learning, deep learning with strong critical thinking and problem solving skills.

EDUCATION:

New Jersey Institute of Technology, NJ May 2020

Master of Science in Data Science

Coursework :

Machine Learning, Deep Learning, Medical AI, Applied Statistics, Data analytics with R, Bigdata, Statistical Method in Data Science, Data Management System Design, Data Analytics With Info Systems Uttarakhand Technical University, Dehradun, India Jul 2015 Bachelor of Technology in Computer Science Engineering SKILLS:

Programming Languages : Python, R, Java, C

Libraries: Pandas, NLTK, NumPy, Keras, TensorFlow, Sklearn Machine Learning: Natural Language Pre-processing, Cluster Analysis, Decision Tree, Linear Regression, K-Nearest Neighbors, Multi- Layer Perceptron, Neural

Network, Naïve Bayes Classification, Random Forest, SVM Deep Learning: Convolution Neural Networks, U-NET, De-convolution Tools: AWS, SQL, MiniTab, Linux, Tableau, RapidMiner, Ubuntu EXPERIENCE:

Insergo Pvt. Technology Limited, Dehradun, India July2015- Jan 2017

Performed Data mining and Data analysis using JMP(SAS).

Performed data mining on dataset which includes descriptive variables of marketing and operational activities of various fast food branch locations, to gain insights on how to improve sales and potentially open new locations.

Ran regression analysis on dataset and used statistical output of the model to make decision on how to adjust operational activities to best manage and operate a fast food restaurant.

Analysed Regression Results and used results of model to estimate expected monthly revenue of proposed restaurant.

Analysed results using neural net methodology and compare the results to the regression analysis for expected revenue. Risk Scoring of Readmissions for ACO patients

Applied descriptive statistics and produced visualization using R to understand the claims data.

Developed predictive model for risk stratifying based on machine learning (decision trees) and probabilistic models (logistic regression)

Developed hybrid predictive model by combining unsupervised techniques to improve accuracy. Internship: ACG Worldwide, Mumbai, India June 2010- Aug 2010

Projected higher dimensions into lower dimensions using PCA, LDA and preserved the data. Random Forest and SVM achieved 85% accuracy.

Also used Bagging and Boosting ensemble methods where the accuracies were better. Implemented interface for Decoding QR Code using Qt with the use of Halcon library. ACADEMIC PROJECTS:

Conversion of clinical MRI scans into research scans

Implemented the U-Net, to convert the clinical to research MRI scans by increasing the dimension.

For evaluation implemented L1 loss to compare our produced output MRI scan to the original MRI. Classification of Restaurants Reviews Using NLP

Analysed the data and performed the necessary data pre-processing.

Implemented NLTK library functions (pot-stammer, count vectorizer) to process the textual data.

Implemented Gaussian and Decision tree models to classify the reviews and achieved the highest accuracy of 86%. ASHRE- Great Energy Predictor 3

Analysed data and applied data pre-processing methods to produce trainable data.

Implemented models; KNN, Decision Tree and Random Forest, to predict energy consumptions for metered buildings and evaluated models by calculating Root Mean Squared Logarithmic Error score. Dimensionality Reduction of Data using Principal Component Analysis

Performed feature selection using Chi-square to reduce the dimension of SNP data and performed classification using Support Vector Machine.

Trained Mini Image Data Set Using Convolution Neural Network (CNN)

Trained CNN model to perform classification of images using Keras and achieved 67% accuracy.

Optimized network using Adaptive Delta optimiser and used batch normalization and dropout to avoid overfitting, which increased accuracy of the model to 76.8%.

Contact this candidate