Publication Date

Spring 2026

Degree Type

Master's Project

Degree Name

Master of Science in Computer Science (MSCS)

Department

Computer Science

First Advisor

Navrati Saxena

Second Advisor

William Andreopoulos

Third Advisor

Jonathan Qu

Keywords

Large Language Models, Retrieval-Augmented Generation, Research Discovery, Graduate Education, Natural Language Processing, Vector Databases

Abstract

When I began my journey towards a master’s degree, I found myself in a daunting situation to find a topic for building a project or writing a thesis on. There are too many ideas out there to pursue, and finding something that would serve my interests as well as contribute to my university’s academic standards overwhelmed me. I kept looking at the previous year’s projects to get some idea, but that was a journey in itself. After struggling for several weeks and seeing my friends struggle with the same problem, I decided this is the exact problem I want to solve. And I decided my project would be to help other students find their projects. I conducted a two-phase user study with SJSU graduate students, which I used to evaluate the problem and the solution. In the initial survey, 23 students participated and rated the topic selection difficulty at 3.91 out of 5 (where 5 = Extremely Hard), and 57% reported changing their topic because they lacked enough information at the start. After releasing ScholarConnect for a week in a small group, I conducted a survey for 10 students who rated the platform at 4.50 out of 5 for advisor matching and 4.40 for topic discovery. I have collected more than 250 master’s project reports and research papers from our university’s Scholar Works website and converted them into text embeddings. I also have a chat-based interface where students can ask questions and receive suggestions for relevant prior projects, details of faculty members who advise in those areas, and university alumni who have worked on similar problems.

Download

Available for download on Saturday, May 22, 2027

Share

COinS