Hi, I'm Nicholas

Our Capstone project: https://saniyalakka19.github.io/#home

We researched improvements to the SPLADE-based Information Retrieval System. Through our research we determined reducing the size of BERT-generated query and document embeddings based on the SPLADE generated term-importance score resulted in significantly lower retrieval time while maintaining Recall and Mean Reciprocal Rank of the unadulterated system, improving overall system efficiency 

I outline a novel approach to document reduction as a preprocessing step for transformer-based model inference and report ROUGE scores for a distill-bart model, trained on the abstractive summarization task.

We created podcast recommender based on two architectures - Matrix Factorization and Deep Neural Network Two-Tower Model - using iTunes podcast data and report results.

We ran a Randomized Control Trial using Prolific, providing retail investors supplemental ESG information as treatment and measuring the change in their propensity to invest in a company as the treatment effect.

We created a dashboard, aimed at retail investors, for exploring the relationship between technical indicators and stock performance, using Tableau. Available here

A statistical analysis of the factors contributing to the popularity of Massive Open Online Courses, with the objective of proving guidance to stakeholders to improve the platform.

A mock research project proposal aimed at the management of Sotheby's. The project focuses on understanding the motivations of NFT art and traditional art investors, to better cross-sell assets to these groups.