A New Chapter

Worked with different teams, including business leaders and product owners, to gather and understand requirements for the AI-Powered Food Inventory Management System. Created clear user stories and acceptance criteria that aimed to reduce food waste from 18% to 10%, potentially saving Boston University approximately $1.3 million annually. Coordinated closely with the development and quality assurance teams during project meetings—like planning sessions and daily check-ins—using Azure DevOps to manage tasks and track progress, while helping to improve software delivery processes, contributing to a projected payback period of 3.9 years for the project.

Employed Naïve Bayesian, Boruta, and CFS feature selection techniques on a large dataset of 300 dependent variables, post preprocessing to ensure and consistency of the data before applying 6 different machine learning classifiers in R. Applied hyper-parameter tuning to classifiers to achieve a True Positive Rate (TPR) exceeding 80% in predictive modeling, showcasing exceptional proficiency in handling imbalanced raw data within a comprehensive data mining framework.

Developed and implemented methodologies for optimizing portfolios, using python, utilizing dividend yield and market volume weighting, which improved risk-adjusted returns by around 15% over a 5 year back testing period.Created a more precise forecasting model with a 20% decrease in portfolio variance by incorporating time series analysis for over 100 tickers`, optimizing portfolio rebalancing strategies and designing an extensive stock data analysis pipeline.

Implemented and automated data operations to Extract, Transform & Load (ETL) an unstructured large 4GB taxi database with 15 million rows of data after data cleaning and validating, into a staging table first and then structured relational database using DDL to eventually create a Data Mart (Dimensional Data) for Online Analytical Processing.

Achieved increase in query performance by applying PL/SQL and T-SQL to create partitions, stored procedures, window functions and CTE to substantially tune and reduce query processing time by 4 to 6 seconds.

I designed a very effective yet straightforward database design in its best normal form as part of the Database Management System curriculum to eliminate any redundancy. This was a semester-long, multi-iteration rigorous project that was completed with the professor's supervision. I was astounded to watch the staff, student employees, and chefs manage the dining establishment where I used to work because I was in charge of checking up on the inventory and usage. For this reason, I chose to do database study and create this project. 

This innovative database design project was tailored to streamline and enhance the user experience in booking hotels and stays. With a focus on user-friendly interfaces and efficient data management, our system aims to simplify the process of finding and reserving accommodations for a relaxing staycation. Through meticulous design and optimization, "BookMyStaycation" promises a seamless platform, empowering users to discover their ideal getaway destinations effortlessly.

I used Python to use the concepts I learned in the Information Structures course to develop this straightforward but engaging game. This game had an obstacle course tailored to the age group chosen, as well as three stages for three important age groups. The hardest part of this job was not using any libraries at all, even though I had done some quite advanced coding that included using machine learning. Everything had to be coded from scratch.

In today's era using GPT for getting answers to any questions of yours has become a go to thing and therefore this was a heavy research between the 2 top leading GPT algorithms. This research compared chat GPT and Bard GPT on various aspects such as the Accuracy, the Promptness, the imagination capability , the aptitude, etc. Finally I was able to conclude on the basis of such metrics the better GPT to believe in.