Machine learning, Deep learning, Data scientist

DHRUUV AGARWAL

1-812-***-**** *********@*****.*** *******@**.***

kaggle.com/dhruuv111 linkedin.com/in/dhruuv-agarwal111 github.com/Dhr11 dhr11.github.io EDUCATION

Indiana University, Bloomington, USA GPA 4.0/4.0

Masters of Science in Data Science Aug 2018-May 2020 Relevant Courses: Deep Learning, Machine Learning, Adv. Natural Language Processing, Search, AI, Applied Algorithms International Institute of Information Technology, Hyderabad, India Graduate courses: Operating systems, Discrete Math & Algorithms Fall 2017 Manipal Institute of Technology, Manipal, India GPA 8.3/10 Bachelors of Technology in Electronics and Communication Jul 2012-May 2016 PROFESSIONAL EXPERIENCE

Image and Text Analytics Intern Siemens Healthineers, Malvern, PA, USA Jun-Aug 2019

• Improved efficiency of image database retrieval by using DICOM text metadata to infer patient image characteristics

• Proposed novel method combined character and word level information to handle unseen words and abbreviations

• The reframed problem statement was executed in Keras, and used a custom combined loss of Cross Entropy and Dice

• Presented and discussed the developments of SOTA work in Language Understanding, XLNet, at Deep Learning Seminar Research Assistant, Machine learning Indiana University, Bloomington, IN, USA Feb-May 2019

• Explored plant gene data to build a model to identify elements in promoter regions, crucial for transcription Systems Engineer Siemens Healthineers, Bangalore, India Jul 2016-Jul 2017

• Worked on TrueD, a part of Syngo Classic, which enabled doctors to compare scans of patients at separate timepoints

• Developed and maintained the application to render the 3D image reconstructed from diagnostic MRI and CT scans

• Led a team of five in initial project to develop a chat console application using Named Pipes with C++ 11 features

• Migrated the entire code base of the TrueD application written in C++, from 32 bit to 64 bit version TECHNICAL SKILLS

• Machine Learning: Bagging, Boosting, Regression, SVM, Markov Model, Dimension Reduction (PCA, SVD), CNN, RNN

• Languages/Database: Python, C++, Java, R, C, MySQL, PostgreSQL

• Libraries: Pytorch, Keras, Numpy, Pandas, Tensorflow, NLTK, OpenCV, Scikit-learn, Lucene, SpaCy

• Software: Visual Studio, Git, MATLAB, Apache Spark, Octave, Hadoop, Docker, AWS, Google Cloud Platform

• Visualization: ggplot2, Matplotlib, Seaborn, Tableau PROJECTS

Generative Dog Images Kaggle, Computer Vision, Pytorch Jul-Aug 2019

• Secured global rank 31/927 in the task of Generating realistic Dog Images with 110 MFiD score on Stanford dataset

• Tweaked Deep Convolutional Generative Adversarial Network with mapped dog breed as conditional info (cDCGAN) Deep Recommender System on Yelp Reviews Text/NLP, Recommendation, Keras Nov-Dec 2019

• Predicted review rating for user-business pair using Convolutional Recommender system with 1.04 RMSE error score

• Identifying important users for a business and vice versa via Attention mechanism in learning user-item embeddings NFL Bog Data Bowl Kaggle, Time Series Forecasting, Regression, Python Nov 2019

• Ranked among top 19%, in Predicting the number of Yards Gained after ball handoff from play tracking data

• Engineered features for orientation, formations etc and then trained Multi-layer perceptron and LightGBM models U-Net Semantic Segmentation Computer Vision, Pytorch Jan-Feb 2020

• Implemented U-Net model from scratch, to identify and detect different objects at pixel level in RGB images

• Achieved 72% Mean Intersection over Union score on the Pascal VOC 2012 dataset with 21 different object classes NBA Outcome Predictor Python, BeautifulSoup, Pandas Jan 2019

• Scraped data from websites and used visualization to construct features to predict match winner based on past data

• Evaluating models like Random Forest and XGBoost, Support Vector Machines gave F1 score of 0.72 on 2017 season Seq2Seq for Abstractive Text Summarization Text/NLP, Keras Nov-Dec 2019

• Generated abstractive text summaries for news articles in Daily Mail dataset, and achieved ROUGE-1 score of 24.1

Contact this candidate