Data Electrical Engineering

TEJASH SHAH

*********@*****.*** San Jose, CA, ***12 669-***-****

linkedin.com/in/tejash-shah github.com/Tejash-Shah EDUCATION

M.S, Electrical Engineering GPA: 3.7

San Jose State University Expected May 2018

B.E in Electronics and Telecommunication

Institute of Engineering and Technology Devi Ahilya Vishwavidyalaya (I.E.T.D.A.V.V) May 2016 SKILLS

Programming language: Python, C

Database and Big Data tools: MySQL, TeraData, Apache Spark (PySpark) Data Exploration and Visualization: Pandas, Scipy, Numpy, Matplotlib, Seaborn, Plotly Libraries: scikit-learn, H2O, TensorFlow, Keras, urllib, BeautifulSoup, OpenCV, NLTK, Scipy, Regular Expression(re) Tools: Jupyter Notebook, PyCharm, Weka, MySQL Workbench, Git, GitHub, Octave, Matlab Mathematical models: Linear, Logistic, SVM, KNN, K-means, PCA, LDA, QDA, SVD, Decision Tree, Random Forest, Gradient Boosting, XGBoosting, Adaboost, Perceptron, CNN, RNN, LSTM EXPERIENCE

Software Engineering Intern – MoboDexter Inc., San Jose Sept ‘17- Dec. ‘17 Tools & Technologies: Python, AWS MySQL, MySQL Workbench, Docker

Developed a machine learning script that fetches the data from AWS MySQL and sends back the prediction to the database and in turn enabling ML capabilities on PAASMER IoT Platform.

Developed a production level ML code using pickle library to run on edge devices like Raspberry Pi.

Added ML module to the existing PAASMER architecture by dockerizing all the modules. Graduate Research Assistant -San Jose State University Feb. ‘17-May ‘17

Developed Emergency communication using MCU CC1350 SensorTag kit and Raspberry Pi to monitor the temperature and pressure of the critical storage section of the production industries

Implemented mesh ne among sensors, connected to Raspberry Pi which in turn connected to Internet PROJECTS

Pover-T Tests: Predicting Poverty Challenge on Drivendata [Python] Feb. ‘18

Implemented ML Regression pipeline involving Imputation, SMOTE, GridSearchCV and XGBoosting to predict a given country is poor or not

Achieved a mean log loss of 0.18191 and a rank of 201 out of 2310 participants. (top 10%) Prediction model for fracture toughness [Python] Aug. ‘17 – present

Predict stress at which material starts to break having different no of pores using Deep ConvNet

Data points are converted into quantized image having pixel values using OpenCV

8 models optimized with GridSearchCV are experimented and evaluated against analytical solutions. CDiscount Image Classification Challenge on Kaggle [Python, Keras] Fall ‘17

Applied pre-trained models like VGG16, VGG19, ResNet50, Inception, VGG19 with real time data augmentation to classify the products into category id based on their images

Handled unbalance in data using SMOTE over sampling leading to increase in accuracy from 37% to 85%

Worked on 58GB training dataset which has 15 million images on GPU Tesla k40 to achieve high accuracy

Selected as the best algorithm team among 10 other competing teams in the academia. Sentiment Analysis of IMDb Movie review [Python] Fall ‘17

Implemented KNN to classify IMDb review of 25000 movies with an accuracy of 85.8%

Vectorizer and cosine similarity functions were used to extract the feature data

Handled HTML artifacts using parsing, removed stop words and punctuation for data preprocessing Classification of Handwritten digit dataset [Python] Fall ‘17

Compared performance of different classifiers such as SWISH function, Linear Classifier, KNN, RBF NN, one two Hidden layer MLNN, SVM, Random Forest, Decision Tree, Naïve Bayes and Logistic regression.

Performed dimensionality reduction (PCA, SVD and LDA) and Regularization (Lasso and Ridge) on dataset leading to improved performance.

Compared using performance metrics like Accuracy, Precision, recall, F-1 score and confusion matrix. Comparison of Routing Protocols [Python] Fall ‘17

Developed a script using Regular Expression(RegEx) to fetch transmission time, reception time, sequence id, source Ip and destination Ip from a IP Trace file.

Determined which protocol is better among DSR, DSDV and AODV by comparing average delay and maximum delay

Distributed Learning Management System using BSD Sockets, API and SQLite [C, Sqlite3] Spring ‘17

Implemented SQLite database on the server side to store files and user credentials

Developed C code for multi-threaded concurrent server to allow simultaneous remote connections

Developed C code for connection oriented client to access the server, fetch files, view course and grades. Network Packet Analyzer using Wireshark, Dpkt and PyGeoIP [Python] Spring ‘17

Implemented PyGeoIP to correlate IP address to physical location of the unauthorized user

Developed Python code for parsing the packets captured in a pcap file using Wireshark

Identified DDOS attack in progress and blocked all users who are downloading from a blacklisted site Threat Alert using IoT and Raspberry Pi [Python] Fall ‘16

Implemented using Raspberry Pi 3 to capture live image of threat/danger situation

Detected harmful objects by using IBM Bluemix Watson Visual Recognition API

Received push notifications on cell-phones within a minute Blind Companion AT&T IoT Hackathon [Python] Sept. ‘16

Established MQTT server to send the speech to clients

Detected visuals and text in the image using Watson’s visual recognition API

Developed Python code for sending the captured image to IBM Bluemix and result to clients CERTIFICATE

Managing Big Data with MySQL by Duke University on Coursera Feb. 25, 2018-present Machine Learning by Stanford University on Coursera Jan. 1, 2018- present Improving Deep Neural Networks by deeplearning.ai on Coursera Dec. 31,2017 Neural Network and Deep Learning by deeplearning.ai on Coursera Dec. 1, 2017

Contact this candidate