I'm planning to apply to universities as a computer science student interested in AI and machine learning. I'd like to be considered for teaching assistant or research assistant opportunities, so I want to build a strong portfolio before applying. Would publishing a research paper help, and what kind of project using AI, machine learning, or LLM technologies would be substantial enough to develop into a worthwhile paper?
3 Answers
A promising direction would be studying documentation drift in LLM-based coding agents. For example, you could measure how an agent’s performance changes when documentation becomes outdated over several sessions, especially when the agent modifies code without updating the documentation. A well-defined benchmark, repeatable experiments, and clear evaluation metrics could turn that into a solid research project.
For a master’s application, a paper can show that you understand the research process, but it isn’t the only way to stand out. A useful, well-engineered project with users, clear documentation, testing, and thoughtful evaluation may be just as impressive. Start by choosing a real problem, then compare your approach with existing methods and explain where it succeeds or fails.
Before choosing a topic, read several recent papers in the area you’re interested in. Pay attention to their limitations, unanswered questions, datasets, and suggested future work. The best undergraduate projects usually extend an existing experiment in a small but measurable way rather than trying to invent an entirely new field. Reproducing a paper’s results and adding one useful comparison can also be valuable.
That makes sense. I’m mainly looking for a direction because I’m not sure what would count as a strong enough topic.