Meet PolyGPT, our AI-powered Project Idea Generator
Tell PolyGPT your favorite subject and a few hobbies you're passionate about to generate personalized project ideas.
Log in to explore PolyGPTExplore project ideas with the help of PolyGPT
See your personalized project ideas come to life with PolyGPT! Log in to brainstorm ideas.
We know the jump from your interests to a research project idea can be tricky.
That's why we've built tools to help you find the perfect project idea to lay the foundation of a successful project. Simply copy your favorite idea to your clip board, and paste it in the project idea section of the application.
Browse project ideas by Polygence mentors
Detecting Bias in Language Models
Test whether an open-source LLM shows bias in hiring, education, or criminal justice prompts.
Philosophy, Neuroscience, Biology, Computer Science, Cognitive
Fake News Classification Model
This project involves building a simple machine learning model that can classify news articles as real or fake based on their text. The student would use a publicly available dataset, learn how to clean and prepare text data, and train a basic model such as logistic regression to make predictions. They would evaluate how well the model performs using clear metrics like accuracy and analyze examples where the model makes mistakes, learning about both the power and limits of AI systems. The final outcome is a working fake news classifier and a short report explaining the model’s performance and insights.
Philosophy, Neuroscience, Biology, Computer Science, Cognitive
Detecting AI hallucinations
Build a simple system that flags potentially hallucinated AI outputs using: (1) Retrieval checks (2) Fact consistency scoring (3) Multiple-model agreement
Philosophy, Neuroscience, Biology, Computer Science, Cognitive
Evaluating the Moral Reasoning of Contemporary AI Systems
Produce a rigorous, reproducible literature review that summarizes how current LLMs and related AI systems perform on moral and ethical tasks, synthesizes evaluation methods and datasets, identifies major gaps, and recommends next research directions and benchmark practices suitable for publication.
Philosophy, Neuroscience, Biology, Computer Science, Cognitive
Testing a New Dark Energy Parametrization with Mock Cosmological Data
Recent large-scale structure data have sparked renewed interest in the possibility that dark energy may evolve over cosmic time. In this project, the student will test the robustness of a newly proposed dark energy energy-density parametrization using mock datasets. The project will involve generating synthetic cosmological expansion data under controlled assumptions, performing Bayesian statistical inference to recover model parameters, and comparing the performance of the new parametrization against commonly used alternatives. The student will investigate questions such as: Does the new model introduce biases? Does it improve flexibility without overfitting? Under what conditions can evolving dark energy be reliably detected? This project serves as a hands-on introduction to statistical inference, model comparison, and the careful testing of theoretical proposals before applying them to real observational data.
Statistics, Physics
Extending CMBverse: Visualizing Secondary Anisotropies of the CMB
The cosmic microwave background (CMB) not only encodes information about the early universe, but also carries subtle imprints from the late-time universe as photons travel toward us. In this project, the student will develop clear visualizations of one or more secondary anisotropy effects — gravitational lensing, the thermal Sunyaev–Zel’dovich (tSZ) effect, and the kinetic Sunyaev–Zel’dovich (kSZ) effect — and integrate them into the CMBverse website. The goal will be to produce high-quality plots that isolate and explain how each effect modifies the primary CMB signal, accompanied by concise, accessible explanations describing what physical processes generate these distortions and what they teach us about dark matter, dark energy, and structure formation. By the end of the project, the student will have contributed new educational research tools to a public-facing platform, while gaining experience in numerical modeling, scientific visualization, and translating technical physics into clear explanations. This project is well-suited for students interested in connecting theory, computation, and science communication.
Statistics, Physics
Building a Stock Price Prediction Model: Can AI Beat the Market?
This project explores whether machine learning can actually predict stock prices better than traditional methods. We will collect historical data on a chosen stock or index, build prediction models using techniques like linear regression, random forests, and LSTMs (recurrent neural networks), then rigorously test their performance. The student will learn Python programming, work with financial APIs, and discover why this problem is harder than it looks. I help students avoid common pitfalls like data leakage and overfitting, and explain results honestly. The final product is a finance + AI research paper with code, suitable for showcasing technical skills to competitive programs.
AI/ML, Economics, Finance
College Application Help
Can design a college application proposal from brainstorming all the way to personal statements. Includes scholarship or fellowship applications! Previous experience in Caltech, MIT, Brooke Owen Fellowship, and various scholarships.
Economics, Engineering
How smart are AI Applications?
Artificial Intelligence (AI) aims to mimic human intelligence. When successful (especially generative AI) it produces applications that can exhibits behavior like what smart people do. Applications such as ChatGPT can engage in a a serious conversation with a human being, come up with a piece of code in Python, Java, C++ or any other program ming language to solve a complex problem, can write a smart power point presentation on a topic can produce videos according to users request and specification, etc. However, though AI application may behave like smart human beings in some areas, the way AI works is fundamentally different from the way humans acquire intelligence. For example, while AI depends crucially on very large amounts of data and on previous encounters of clues in its data (machine learning) to make decisions, human intelligence can make decisions on unseen and novel problems very easily. This research is multifaceted and different students can different aspects of the general problem under investigation.
AI/ML
Artificial Intelligence and Hunan Intelligence
This is an investigation into the relationship between human intelligence (HI) and current applications of Artificial Intelligence (AI). We examine the basic assumption that AI is an attempt to mimic natural HI. We want to determine what areas both AI and HI agree and what areas they disagree. We would like also to consider how far AI reflects our understanding of what Human Intelligence is. In areas where they disagree, what impact does that have on current AI applications.
AI/ML