Inspiration
We wanted to build a student-focused AI assistant that feels like calling a real advisor. Students deal with finances, scholarships, housing, nutrition, and confusing documents — all at once. Instead of juggling tabs and paperwork, we envisioned one voice agent that listens, answers, and remembers.
What it does
Advisr.ai is a voice-first AI agent that students can call to manage almost anything. It answers questions, generates concise summaries after every conversation, and stores them as a permanent record. Students can upload documents — like leases, tuition statements, or scholarship requirements — which are indexed and analyzed using vision-based RAG, letting the agent explain dense PDFs and images in plain English. Beyond Q&A, it can build budgets, meal plans, grocery lists, and reports, making it a true one-stop advisor.
How we built it
We built the backend in FastAPI with Pydantic models to keep data structured and consistent. ChromaDB handles vector storage and semantic retrieval, while our vision-based RAG pipeline processes both text and images from uploaded documents. Gemini powers multimodal summarization, conversation recaps, and document analysis, while Perplexity provides web-grounded answers for fresh information. A custom database manager stores users, finances, tuition data, documents, and chat recaps, giving the agent memory across calls. The document service manages uploads, indexing, and search, tying everything together into the RAG workflow.
Challenges we ran into
One challenge we ran into came from deployment. We ran into small issues around document handling and environment setup when moving from local to hosted deployment, but resolved them so the project could run live.
Accomplishments that we're proud of
We are proud that we deployed this project successfully, not just as a local demo. Advisr.ai is running live, end-to-end, with real APIs, document uploads, vision-based RAG, and conversation summaries. Also, we are proud that we built a working voice agent that can handle any student problem in one weekend. It manages voice calls, uploads and analyzes documents, grounds answers in indexed knowledge, and keeps a permanent record of every conversation. Instead of a chatbot demo, Advisr.ai acts like a true assistant with memory, clarity, and practical utility.
What we learned
We learned that retrieval and generative AI together make a much stronger experience than either alone. Summarization is critical — students value short, bullet-point recaps more than raw transcripts or long paragraphs. We also saw how a single system can flexibly cover many different scenarios, from finances to nutrition to compliance, if the foundation is strong.
What's next for Advisr.ai
Next, we want to expand Advisr.ai with full calendar and email integration, so summaries automatically generate reminders. Our goal is to turn Advisr.ai into a platform for student life management — one assistant that listens, remembers, and advises across every aspect of college life.
Built With
- chromadb
- colpali
- fastapi
- gemini
- nextjs
- perplexity
- postgresql
- pydantic
- python
- rag
- react
- tailwindcss
- typescript


Log in or sign up for Devpost to join the conversation.