BrinRank

Inspiration

Alex initially considered selecting a research project that would have been excessively challenging for him. The previous year, his ambition had prompted him to pursue a particularly demanding endeavor; however, its complexity ultimately prevented him from making meaningful contributions to the professor's work.

What it does

BrinRank is an AI-powered tool that transforms overwhelming research papers into accessible learning roadmaps for undergraduates and self-learners. Users simply paste an ArXiv ID, and it leverages LLMs (like Claude) to extract 3-5 key concepts from the abstract, then builds a prerequisite knowledge graph—mapping out sequential dependencies (e.g., "Learn RNNs before Attention Mechanisms"). Using basic graph algorithms (BFS for shortest paths), it generates 1-2 tailored learning paths with tips and resource links, visualized in a clean UI as cards and simple flowcharts. Pre-built examples for classics like Transformers, ResNets, and GANs ensure reliability, making dense papers feel like a guided syllabus rather than a wall of text.

How we built it

BrinRank was developed as a web application, starting with the conceptualization of easing prerequisite navigation for undergrads via AI-driven graphs and summaries. The team selected a modern stack: React 18 for modular UI, Vite for rapid builds, React Router for navigation, ArXiv and Google Gemini APIs for data/AI extraction, and React Force Graph for interactive visualizations. Initial setup involved cloning a Vite React template, installing dependencies, and configuring a .env for the Gemini API key, followed by launching the dev server. Core implementation focused on ArXiv search, recursive Gemini prompts for concept/prereq extraction, and an algorithm for weighted shortest paths by study hours. Finally, the project was documented with usage guides, troubleshooting, and MIT licensing.

Challenges we ran into

Our biggest challenge was the 12-hour time limit. API rate limits on Claude threatened live demos. Minor bugs arose when graph outputs weren't compatible with the UI. As a beginner-heavy team, finding out prerequisites accurately for complex papers like GANs and understanding key concepts was difficult to get used to.