Professional Documents
Culture Documents
Projects
XYZurl.me: (Chrome Extension, WebApp)
‣ A web application and google chrome extension to shorten URL’s. Implemented urlTracking using reverse-geocoding, using IP information.
‣ Web shortened URL is stored in MongoDB, while Redis instance is used to cache chrome-extension generated URLs for a 3 hour period.
‣ WebApp is implemented in NodeJs and hosted on Heroku Cloud. Facebook login is implemented using PassportJs
Tweenalyzer: (WebApp)
‣ Twitter data and sentiment analysis WebApp deployed on Heroku Cloud using Flask Framework(Python)
‣ Tweepy library is used to fetch streaming tweet data, Google charts is used for charting and Pandas is used for data cleaning.
‣ Dasboards include Sentiments PieChart, GeoChart with tweet location & sentiment gradient tooltips, tweet language analysis, retweets table, etc
Talkativ: (Github)
‣ chatroom app built using Express framework (NodeJs), socket.io, Redis, MongoDB, nginx & Facebook login (PassportJs)
‣ Users can create topics and host / join group chats. Users can also view actively online and available members in the chatroom.
Dev-Support: A marketplace for developers to showcase & sell projects /code (Documentation)
‣ Presented at hackathon HackEd 2017 | Skills: Python, Github API , Stripe API, Materialize
Safari Rush AI: An AI to solve any puzzle configuration for the game of Safari Rush Hour (Documentation and Code)
‣ Uses Simplified Memory Bounded A*algorithm with blocking vehicle heuristic
‣ Decreased search time by 30% by using a combination of memorization, zobrist-hashing and priority-queues
Ataxx AI: AI Bot that plays optimal moves in constrained time to win the game of Ataxx, (Documentation and Code)
‣ Uses Alpha-beta pruning with iterative deepening
Traditional Chinese Medicine Image Classification (Github Writeup)
‣ K-Means clustering algorithm is used to accurately differentiate between the foreground(Herb), from the background.
‣ A python script is written to automate extracting and labelling the target herbs from the given images using segmentation.
‣ The segmented images were trained using a Random Forest Classifier. 98% Classification Accuracy is achieved
Improving image classification accuracy of Kaggle Cat vs Dogs dataset using Deep-Learning (Github Writeup) , (Video Demo)
‣ Increased classification accuracy to 91.18% | Skills: Keras(Theano) , Python , OpenCV
‣ Performed just-in-time Image Augmentation to tackle overfitting on the reduced Cats Vs Dogs dataset.
‣ Utilized Random Search algorithm as a base for effective hyperparameter tuning.