KAPIL C. KHOND

312-***-**** addgmq@r.postjobfree.com linkedin.com/in/kapilck github.com/kapilkhond EDUCATION

Illinois Institute of Technology, Chicago May 21

Master of Data Science GPA: 3.5

: Coursework: Applied Statistics, Data Preparation and Analysis, Big Data Technologies, Deep Learning, Natural Language Processing

University of Mumbai, Fr. Conceicao Rodrigues College of Engineering June 15 Bachelor of Engineering in Information Technology (Graduated with First Class) PROFESSIONAL SKILLS

Programming Languages: Python, R, Java, SQL, UNIX Shell Scripting Big Data Technologies: Hadoop, Apache Pig, Apache Spark, Hive Machine Learning Libraries: Pandas, Numpy, Scikit-learn, NLTK, Matplotlib, Tensorflow, Keras Web Technologies/Frameworks: RESTful API - Flask & Pyramid Databases: Oracle, MySQL, MongoDB

Data Science Skills + Engineering: Machine Learning, Probability & Statistics, Linear Algebra, Web Scraping, Data Integration, Data Wrangling, Data Analysis, Data Visualization, Inferential and Descriptive Statistics, Data Exploration, Data Mining, Statistical Modeling, Business Intelligence, ETL WORK EXPERIENCE

Data Scientist, ZOC Technologies Pvt. Ltd. (client Mahindra and Mahindra Financial Svc. Ltd.) Mar 19 - Jul 19 Credit Risk Modelling

• Predicted the probability of a borrower defaulting on the loan thereby reducing the overall losses by 5%.

• Created risk bands based upon risk score and assisted with the design and development of policies based on such risk bands.

• Analyzed data to identify trends, patterns, insights and discrepancies using Python and conveying ideas with visuals created in Matplotlib, Tableau. Removed noisy and irrelevant features using feature selection techniques.

• Developed a supervised classification model using XGBoost (implementation of gradient boosted decision trees) and random forest on selected features in Python.

Data Scientist, The Zero Games Pvt. Ltd. May 17 – Feb 19 User-Profiling Engine

• Predicting user’s persona (user interests, gender, socioeconomic class), thereby helping in behavioral-based ad- targeting.

• Developed web scraper engine with the help of Beautiful Soup (Python library) and Selenium WebDriver to collect raw data (around 0.2 million records) from the web.

• Pre-processed the raw data using Natural Language Processing techniques like removing stop words, Tokenization, stemming, word segmentation and implemented TF-IDF and Text Rank algorithm for keyword extraction whose result will be used as features for the model.

• Built Artificial Neural Network for classifying users into different personas using TensorFlow which resulted in an increase in the Clickthrough rate (CTR) from 0.6% to 0.96%.

• Developed Restful web service APIs in Python (Flask web framework) along with RabbitMQ to predict user persona in real-time for providing production ready solution. Backend Developer, The Zero Games Pvt. Ltd. Aug 15 - Apr 17

• Designed ad-server prototype (using the Holt-Winter forecasting algorithm and Linear Programming in R).

• Developed data-crunching applications in Python for ETL processing of the data coming to the ad-server.

• Deployed a new back-end API of analytics platform which improved latency by 12% by designing and implementing Restful web service APIs in Python (Pyramid Web Framework). PROJECTS

Analyzing Music Tracks for Popularity Prediction

• Exploited pig scripts to clean and load the complete data into hive tables.

• Used hive scripts for exploratory data analysis on the song’s features for summarizing and interpreting hidden information.

• Predicting the popularity of a song using logistic regression (machine learning algorithm) in PySpark with F1 score - 0.84.

Generating Optimal Portfolio by Analyzing Stock Indicators

• Using multiple technical indicators to predict the most probable stock returns over time.

• Prepared the data by handling missing values, resampling it to weekly timeframe and adding 68 technical indicators as features and a response variable being log return.

• Performed time series analysis using ARIMA and Developed LSTM model using Keras to predict the log returns.

Contact this candidate