Nishant Revur

▪ Estimating the number of views, likes and hits a new song would get on Spotify and YouTube
▪ Mining social media cues such as sentiment towards artist, past songs, their reach and latest music trends
▪ Clustering music tracks by features such as danceability and valence to unearth patterns
▪ Boosting will be used to reduce bias and stand out

▪ Automatically tags new questions with relevant topics based on title and content
▪ TF-IDF along with time series analysis selects most popular tags and weights them by upward/downward trend
▪ Bigram and trigram analysis provides additional insight into words that occur together
▪ Convolutional Neural Network to do final tagging of questions

▪ Used python with no external libraries such as NLTK to classify hotel reviews as Positive/Negative sentiment and True/Fake for its genuineness
▪ Implemented a Naive Bayes solution with smoothing to handle unseen data after training
▪ Achieved a 91.7% accuracy on the given data set

▪ Developed a part of speech tagger in python which could tag any language without the use of any external libraries
▪ The solution was implemented using Hidden Markov Models for learning and the Viterbi algorithm for decoding
▪ Attained an accuracy in the range of 88-92% when tested on English, Chinese and Hindi inputs

▪ Developed a cost effective vehicle speed detection program in MATLAB
▪ Engineered the solution using blob detection, background subtraction and optical flow analysis

▪ Modelled an assembler for the SIC/XE instruction set
▪ Incorporated all aspects of assembly language such as labels, retadr, conversion of mnemonic opcodes and data constants

▪ Developed an inference engine in first order logic using backward chaining by contradiction ( resolution )
▪ Takes CNF sentences as input and decides whether given queries are true or false and adds inferred sentences into the knowledge base
▪ Built in loop detection ensures the algorithm isn’t stuck in an infinite loop

▪ Programmed an AI agent to play and maximise chances of winning a game similar to candy crush
▪ Implemented it using minimax algorithms with alpha beta pruning and move ordering to reduce search space and meet strict time constraints

▪ Designed a comprehensive quizzing application to replace the college’s current quiz portal
▪ Offered a diverse range of topics (academic and non-academic)
▪ Constructed statistics for each user such as win percentage per topic, overall and most selected topics and represented it visually using JGraph

▪ Built a carpooling web app for students to ease daily commute, reduce cost and make the travel more eco-friendly
▪ Groups students based on home location, college location and schedule