Projects

Built an android app and a web application that supports real time stock updates, provides visual analytics of indicators and other functionality such as Facebook sharing and favorites.
Languages / Frameworks used : Angular, TypeScript, Node, Bootstrap, HTML5
Github Link : https://github.com/shrev/CSCI571-Web-Technologies/tree/master/HW4
Web Application Link : http://st-env.us-west-1.elasticbeanstalk.com/

This project stemmed from the need for a smarter solution to Bangalore's traffic issues. The architecture includes a deep Q-network traffic signal control agent, with the action-value function modeled as a deep convolutional neural network. The network was trained using reinforcement learning in a traffic microsimulator, SUMO, on an isolated intersection. Algorithm was aimed at reducing average waiting times and queue lengths at junctions. This project also reached the top 30 out of 1000+ teams in the 'Flipkart-Gridlock Hackathon'. My role involved data generation, integration and traffic simulation.
Languages / Frameworks used : Python, Simulation of Urban MObility (SUMO)
Number of Members in team : 3
Github Link : github.com/shrev/Reinforcement-Learning-Traffic-Agent.git

Inspired by the RecSys challeng of 2016, a group of us built a recommendation engine that uses clustering combined with deep learning and collaborative filtering to exploit the implicit data of user interactions from social networking sites in order to recommends jobs. My contributions include coding the deep neural network architecture and testing.
Languages / Frameworks used : Matlab, Python
Number of Members in team : 3
Github Link : github.com/shrev/Deep-Job-Recommendation-for-Career-Oriented-Social-Networking-Sites

Developped a learning machine that continually learns and self-adjusts in order to detect SQLI and XSS in incoming packets. The built system has two modes - Learning and Detection and SVM was used as the primary algorithm to achieve this. I was tasked with backend.
Languages / Frameworks used : JEE
Number of Members in team : 2
Github Link : github.com/shrev/Intelligent-System-for-SQLI-and-XSS-detection

Survey in college conducted to determine people's choice of restaurants and various parameters associated with it. Predicted most preferred restaurant by training a neural network. Implemented a basic feed forward back propagation neural network.
Languages / Frameworks used : C with OpenMP
Number of Members in team : 3
Github Link : github.com/shrev/Localised-Restaurant-Preference-Prediction-using-Parallelised-Neural-Networks.git

This was a self-learning project under Dr. Balaraman Ravindran. This project proved to be my induction to machine learning. I trained different models using AFINN list and the word2vec toolkit. Simple ML algorithms such as kmeans, LR,SVM , naive bayes classifier and random forests used for predicting the sentiment of movie reviews. Results of each classifier compared to determine best suitable classifier for the task.
Languages / Frameworks used : Python, R
Github Link : github.com/shrev/Sentiment-Analysis.git

A stand alone software to simulate timetable choosing for students of universities that follow the flexi credit system.The system built is real time and supports synchronous tasking.
Languages / Frameworks used : Java, Dropbox