Task: Analyze different approaches used for n-gram word predictions and optimize them for resource-constrained settings.
Models analyzed: Analyzed the Hidden Markov model, the Back-off model, character-based LSTM model, and word-based LSTM model.
Inference: word-level LSTM is optimum in terms of complexity and space-efficiency compared to others.
Github link for the code and more details here.

The study mainly consisted of the following three topics:

• Playlist Generation using recommendation systems
• Introduction to Reinforcement Learning for online advertising in Recommendation Systems
• Diving into Reinforcement Learning and how it works

Github link for the code and more details here.

These projects were implemented in Python under the course lecture of Artificial Intelligence (AI):

• Implemented Pac Man game where the Pacman agent tries to find the path through the maze to reach a destination and to locate the food efficiently.
• Implemented search algorithms like BFS, DFS, Uniform Cost Search, A* search with custom heuristics, Minimax algorithm, alpha-beta pruning.

Github link for the code and more details here.

Analyzed the data of Costello’s ACE hardware retail store chain with the primary goal of providing the store with insights that will help them grow their business. It also included performing the following techniques:

• Market-Basket Analysis using the Apriori algorithm: For learning association rules.
• Time series analysis for Product inventory management: To predict the stock of products in the future.
• Loss Leader Analysis: Discounting technique to boost the sales of particular items.

Github link for the code and more details here.

Implemented the Naive Bayes algorithm from scratch for spam filtering.
The dataset used is a subset of 2005 TREC Public Spam Corpus. It contains a training set and a test set.

Github link for the code and more details here.

Spring MVC Restful application for Pizza Delivery System: This project was created during my initial days at eQ Technologic in a group of four.
It was continuously reviewed with respect to Database Design, sequence & class diagram, UML Design, OOP, Flexibility and Scalability.
Due to NDA, code and other details regarding the project cannot be made public.

Performed information extraction based on multiple datapoints from html and PDF documents for Broadridge Financial Solutions as part of advance project.