Rabbit Hole

Inspiration

Every college STEM student knows the dread of staring at a blank screen, trapped by "research paralysis." Students are constantly told they need real-world portfolio projects to land internships, but trying to start one usually means wasting hours Googling tutorials, hunting for clean datasets, or configuring environments before writing a single line of code. We built Research Rabbit Hole to eliminate this upfront friction entirely—transforming a student's major, personal hobbies, skill level, and time budget into an instantly executable, resume-ready passion project.

What it does

Rabbit Hole is an AI-powered platform designed to turn college students from passive learners into active builders. We ask the user to input their major, personal interests, skill level, and time budget. The app instantly pulls a structured, anti-research project blueprint from our database with pre-vetted tool stacks, mock data sources, and pre-formatted XYZ resume bullets. We integrated real-time web search via SerpApi for our mentor to reference for accurate student assistance with up-to-date documentation, API lookups, debugging, and co-designing custom projects from scratch.

How we built it

Gemini, Serpapi, Claude, Grok.

Challenges we ran into

We burn out from the constant prompting but credits burn out faster. AI models default to lazy, generalized advice when providing the user with instructions. Other challenges were tying together Flask routes, SQLite data models, and asynchronous LLM calls under tight hackathon time constraints while maintaining a snappy user experience. Merge conflicts were also a big issue as we built the front end and back end branches were built separately from the start to prevent clashing work. SerpApi has very constrained credit usage under the free plan.

Accomplishments that we're proud of

Achieving instant project loading by leveraging structured local database retrieval paired with dynamic progress tracking. Successfully wiring SerpApi into our mentor layer so students receive accurate, current documentation instead of outdated training data. Creating a seamless workflow where a student stuck on a specific task can broadcast an SOS straight to the community pre-tagged with their exact project blueprint and task ID.

What we learned

Students don't want an overwhelming list of options; they want one fast, reliable path forward. Eliminating decision fatigue drastically reduces friction and increases execution rates. Hiding advanced project phases behind clean drop-down accordions keeps students from getting overwhelmed by long-term scope creep. API integration for SerpiApi and Gemini API to have up to date data retrieval and to solve the credit usage, we had it run searches only on specific problems. We solved generalized AI suggestions by implementing strict "anti-research" negative constraints and enforcing structured JSON schemas so the backend always prescribes exact tools and shortcuts.

What's next for Research Rabbit Hole

Setting up a public domain for users to access from their browser. Expanding the majors list and selectable hobbies to open blanks for the user to have more freedom. Building ingestion pipelines to continuously curate trending open-source APIs and public datasets into our database. Integrating group projects in the community dashboard to help students form micro-teams for larger Gold-tier capstone projects.