Satyajit Kamble

Masters's Student At the Oregon State University · Printable Resume PDF
Email:

I am a master’s student majoring in Computer Science at OSU. I’m a curious programmer who absolutely loves to play with data. Out of fun, I have been building various projects in the domain of deep learning, machine learning and have also been fortunate enough to have worked on a research project in NLP with Dr. Aditya Joshi. I am really excited about the fact that analysing data on varying scales is capable of creating a huge impact within societies and businesses.

I learnt everything I know from the Internet! My learning caps are always on! I am passionate about coding and writing scripts just like my most recent one to Scrape Linkedin profiles and discover top companies (based on ELO score) allowing students to get internship details for a better shot at finding their best fit. I founded and led the Somaiya Machine Learning Research Association (SMLRA) at my undergrad school, which grew to 250+ members including students and faculties. Apart from coding and ML, I fool around, watch movies, make videos and trek a lot :-)

Here’s my favourite quote: “Whatever the mind of man can conceive, it can achieve”.

Education

Oregon State University

Master's Degree · Majoring in Computer Science · GPA N.A/4

---Courses Yet to Start (last update 07/16/2019)---

Sept 2019 - Present

University of Mumbai

Bachelor of Technology · Majored in Computer Engineering · CGPA 7.30/10

Analysis of Algorithms · Data Structures · Neural Networks and Fuzzy Logic · Machine Learning · Artificial Intelligence · Digital Signal & Image Processing · Object Oriented Programming Methodology · Discrete Structure and graph Theory · Advanced Database Mangement Systems

Jun 2015 - Jul 2019

Experience & Research

Directed Research Collaboaration

Data61, Commonwealth Scientific and Industrial Research Organization (CSIRO)
  • Collaborated with an NLP researcher on hate-speech detection from code-mixed data on social media. Extracted 255k+ domain-specific tweets using REST API and created an algorithm to pre-process the data.
  • Trained domain-specific word embeddings to capture semantic subtleties. Designed and implemented CNN-1d, LSTMs and Bi-LSTMs. Evaluated their performance against state-of-the-art statistical classifiers.
  • Results showed a 12% improvement in F-score on a benchmark dataset. This research project resulted in a paper which got accepted at ICON 2018, a national level A-star NLP conference in India.
Jun 2018 - Aug 2018

Undergraduate Research Assistant

University of Mumbai
  • Led a team of 5 and built a question answering chatbot from a given text corpora. Developed the seq2seq encoder-decoder architecture to learn context vectors from training data. Implemented a greedy search decoding module.
  • Improved performance by incorporating the global attention mechanism to allow for refined context discovery.
Aug 2017 - Oct 2017

Projects

LinkedIn Profile Scraping

view tutorial
  • This project was built to aid students like myself to allow them to discover unknown companies and their rating (ranks based on ELO score and Naive score) with respect to other companies.
  • Student and internship/job data is extracted to allow students to easily search company roles/positions and their associated responsibilities, allowing students to save valuable time.
Jun 2019

Skip-gram Model for Word2Vec

View Github Code
  • Led a team of 3 to explore the application of the CBOW and skip-gram model.
  • Scraped textual data from the web and built the skip-gram model (in tensorflow) to find correlations between movies and TV series along with their characters. Also, implemented noise contrastive loss.
Jan 2018

LSTM Stream Sequence Counter

view github code
  • A toy project which used LSTM cells to calculate the number of 1's in an auto-generated binary element dataset.
  • The toy code was primarily to explore and understand how the temporal dynamics of the dataset with long sequences are captured by the modified RNN cells
Dec 2017

Neural Image Dual-Style Transfer

view github code
  • Built a model to transfer the style of 2 images into a third base image. Used the VGG-16 model for transfer learning.
  • Developed a novel approach to combine associated style loss and content loss. Also, incorporated regularization.
Oct 2017

Sentiment Analysis in Game Reviews

  • Created a model for analyzing sentiments of game reviews using statistical classifiers - Random Forests and SVMs. Extracted and utilized several feature vectors such as word n-grams, character n-grams, negation words etc.
  • Improved task accuracy to 92% by using ensemble deep learning models accompanied with GloVe embeddings.
Jul 2017

CNN for CIFAR-10 Using Batch Normalization

view github code
  • Built a ConvNet with batch normalization to classify distorted RGB images in the CIFAR-10 dataset. The primary purpose was to analyze the effect of batch normalization on the inference.
  • This was a personal project to accompany my understanding when I was a part of a team where we built the Siamese Network for facial recognition and verification for an attendance management system.
Jul 2017

Neural Networks for Prediction & Detection

  • Employed NNs for tasks such as:
    • Created a model for tracing trends and predicting stock prices using Gated Recurrent Units (GRUs) and Convolutional Neural Networks.
    • Led a team of 4 for breast tumor detection and classification using deep-CNNs.
    • Built a toy project which analyzed facial landmarks to keep track of blinking.
Oct 2017- Mar 2017

Tensorflow Workflow

view github code
  • I founded and lead the Somaiya Machine Learning Research Association (SMLRA) in my undergrad year, which grew from 3 members to 250+ members in one year. The purpose of this project was to serve as a rudimentary introduction to deep leanring and using the Tensorflow framework, for students/members of SMLRA.
  • Implemented and documented(on Github) various network architectures such as ANN, CNN, LSTM and logistic regression in Tensorflow over the basic MNIST dataset.
Feb 2017

Skills

Contact Info

Mailing address: 218 · 21st NW Street · Corvallis · OR 97330

Email: satyajit[DOT]k[AT]somaiya[DOT]edu

Phone: +91-9757220499