KARTHIK KUMAR PADTHE

*** ******* ** *******, ** *8202

734-***-**** acv7th@r.postjobfree.com

EDUCATION

Wayne State University December 2016(expected)

M.S. in Computer Science(Thesis)

Advisor : Dr. Chandan Reddy

Overall GPA: 3.85

Jawaharlal Nehru Technological University-Hyderabad May 2012 B.Tech. in Computer Science

Overall GPA: 3.00

TECHNICAL STRENGTHS

Computer Languages Java, Python, PHP

Statistical/Machine learning Packages R, MATLAB, Scikit-learn, Pandas, Numpy, WEKA Databases MySQL, MongoDB

Web Technologies HTML, jQuery, XML, JSON

Tools LATEX, Eclipse IDE

Operating Systems Linux, Windows

RESEARCH EXPERIENCE

Wayne State University August 2015 - Present

Graduate Research Assistant Detroit, MI

Consensus Regularized Selection based Prediction : Worked on an ensemble based approach which uses a sequence of non-convex regularizer based Regression models to handle feature sparsity in data. The model was tested on EHR data and synthetic data. Evaluation metrics like AUC, MSE, R-squared and P-value were used to evaluate model performance.

Feature Grouping using Weighted ‘1 Norm for High-Dimensional Data : Worked on a regression model which uses a weighted ‘1 convex regularizer to handle feature grouping in data. The model was tested on 20-newsgroups, gene-expresion data and synthetic data. Evaluation metrics like AUC, MSE, R-squared and P-value were used to evaluate model performance.

Top-N Recommender System using Weighted ‘1 Norm : Worked on a matrix factorization method which uses a weighted ‘1 regularization penalty to minimize the matirx reconstruction error and handle sparsity in the data. The model was test on real world datasets like net

ix, movie lens, yahoo!movies, lastFM, book-crossing and delicious. Evaluation metrics like Hit Rate(HR) and Average Reciprocal Hit Rate(ARHR) were used to measure the performance of the model.

Readmission Analytics on Claims Data : Worked on 5 years claims data with membership, provider and patient claim information. Used state-of-art machine learning models to predict risk on readmission for heart failure patients. AUC was used to evaluate the prediction performance of the models. Wayne State University May 2015 - July 2015

Student Assistant Detroit, MI

Worked on EHR data related to heart failure and cancer patients.

Implemented data preprocessing methods such as feature selection, data normalization and data imputation using R.

Used R based package glmnet to implement di erent regularized regression based models.

Used Java based WEKA libraries to implement state of art Machine learning models. PUBLICATIONS

Ping Wang, Karthik K. Padthe, Bhanukiran Vinzamuri and Chandan K. Reddy. \CRISP: Consensus Reg- ularized Selection based Prediction", In Proceedings of the ACM Conference on Information and Knowledge Management (CIKM), Indianapolis, IN, October 2016. Acceptance rate: 28.88%.

Bhanukiran Vinzamuri, Karthik K. Padthe, Chandan K. Reddy. \Feature Grouping using Weighted ‘1 Norm for High-Dimensional Data", 2016, submitted.

PROFESSIONAL EXPERIENCE

M2M innovations LLP June 2013 - November 2014

Software Engineer Bangalore, India

Involved in writing complex SQL queries to retrieve data from MySQL.

Involved in development of Java based middleware and web application for a IOT related project.

Worked on development PHP based API which returns a JSON on request.

Was responsible for setting up Memcached on Amazon Linux system.

Worked on AWS API to access static content stored on AWS S3.

Was involved in developing user friendly Android based mobile application and connecting to server side API for data exchange.

ACEDAMIC PROJECTS

Semi-supervised learning using frequent itemset

and ensemble learning September 2015 - December 2015

Apriori algorithm was used to identify frequent itemset in labled and unlabled data.

Machine learning models Naive Bayes, Random Forest and SVM were used to form ensemble of models.

Majority voting based procedure was used to make nal prediction.

UCI SMS spam collection dataset was used to test the model.

Data preprocessing methods like stop word removal, stemming were used.

Metrics like F-measure, Accuracy, Precision and recall were used to evaluate model performance. Online Book Store January 2015 - April 2015

A e-commerce web application was developed to host a online book store.

MongoDB was used as the database to store all the information related to books and authors.

MapReduce and aggregation queries in MongoDB were used process user request.

User access behavior was logged on CouchDB for future analysis. Open directory website January 2015 - April 2015

A web application was developed to host a open directory website.

MySQL was used for storing information related to webpages, reviews and their categories.

Worked with complex join commands to retrieve relevant data.

Contact this candidate