Through the use of reinforcement learning techniques, we successfully constructed a Q-learning agent and a Deep Q-learning Network agent which successfully solved the Cartpole-v0 problem in the OpenAI gym toolkit.

Implemented the PageRank algorithm in Hadoop MapReduce framework and Spark using RDD APIs, in order to parallelize computations of ranks, and deployed onto AWS EMR cluster.

Designed and implemented Full Stack application using React, Mongoose, Express and Node.js (MERN), allowing users to register for, and add/search for recipes with RESTful web services/APIs depending on user role.

Vertical Search Engine

Developed custom indexer, retrieval/evaluation system and vertical search engine using Python for increased precision on retrieval models such as Okapi BM25, TF-IDF and vector space cosine similarity algorithms using Python, ElasticSearch, and Kibana.

A project that analyzed Pokémon data to build the dream team for beating Generation 1 Pokémon games. Additionally, the analysis was furthered to build a classifier using Random Forests and Decision Tree to determine whether a Pokémon is of a certain type (legendary).

Developed and deployed a full stack web application using MEAN on AWS Amplify that interfaces with Fitbits via REST APIs to log stress times using indicators (heartrate, EEG,..) and view stress indicators on a dashboard.

Nucleus is a peer-to-peer image sharing Android application that uses an optimal lossy image compression algorithm to reduce the size of images with minimal loss in the quality, and shares this image over a WiFi network using WiFi Direct technology.