www.linkedin.com/in/saurabh-parkar www.github.com/saurabh-parkar Education
Northeastern University, Boston, MA Sep 2019 - Present Khoury College of Computer Sciences
Candidate for Master of Science in Data Science, GPA: 4.0/4.0 Related Courses: Introduction to Data Management and Processing, Supervised Machine Learning and Learning Theory, Unsupervised Machine Learning and Data Mining
K. J. Somaiya College of Engineering, Mumbai, India Aug 2015 - June 2019 Bachelor of Technology in Computer Engineering (Hons), GPA: 8.42/10 Related Courses: Data Warehousing and Mining, Artificial Intelligence, Neural Networks, Data Structures, Advanced Database Management Systems
Programming Languages: Python, R, MATLAB, Java
Machine Learning Intern K J Somaiya College of Engineering, Mumbai Dec 2017 – Jan 2018
Developed a software for facial expression detection using TensorFlow and OpenCV that could detect 7 basic emotions (happy, sad, anger, contempt, disgust, fear, surprise).
Dlib was used to extract 68 facial landmarks which were used to train a deep learning network for classification resulting in 95% accuracy.
Music Genre Classification Sept 2019 – Dec 2019
Developed models to classify music files (.mp3 format) into 8 different genres using 2 approaches- o Processed the music files and transformed them into spectrograms (Mel spectrogram and MFCC). Used the spectrograms as images to train CNN and C-RNN.
o Extracted advanced audio features using LibROSA library and developed models to classify music using classical machine learning techniques.
Multivariate Time Series Forecasting for Energy Consumption Prediction Oct 2019 – Dec 2019
Cleaned the data and performed in-depth exploratory data analysis to understand how different geographical and weather conditions affect energy usage and analyzed their trend and seasonality.
Implemented models to predict Energy meter readings using traditional time series forecasting methods, neural networks and gradient boosting algorithms and achieved an optimal RMSE of 0.79. Visual Image Caption Generator using Deep Learning Oct 2018 – March 2019
Built models to generate semantically correct captions for any given image using a combination of Convolutional and Recurrent Neural Networks resulting in a BLEU-1 score of 0.51 and BLEU-2 score of 0.34.
VGG16 and Inceptionv3 were used for extracting features and LSTM, GRU were used for framing sentences along with GloVe word embeddings. Four different models were built using a combination of above 4 architectures and compared.
Prediction of onset of Diabetes Sep 2018 – Nov 2018
Cleaned, visualized and analyzed the data to understand the relationship between various features and target.
Developed simple models (on a subset of best selected features) to predict the onset of diabetes using Logistic Regression and Random Forest Algorithm, resulting in 85% accuracy. Interests / Activities
Published a research paper “Visual Image Caption Generator Using Deep Learning” in Elsevier-SSRN, April 2019 and presented at the ICAST 2019, held in Mumbai, India.
Second Runner-up at the national level project competition, Prakalpa 2019, held in Mumbai.