Inspiration
We were inspired by the constant challenge students face in navigating the overwhelming amount of information at university. Between scattered departmental websites, confusing class schedules, and important deadline reminders buried in emails, students often waste precious time just finding answers. We wanted to create a single, intuitive, and always-available point of contact—a digital assistant that could bridge the gap between students and the complex university ecosystem, making student life simpler and less stressful.
What it does
UniBot is an intelligent chatbot designed to be a student's personal campus companion. It can:
Answer FAQs: Provide instant answers about tuition fees, library hours, exam schedules, and faculty contacts.
Find Resources: Direct students to mental health services, career counseling, tutoring centers, and club information.
Course & Schedule Management: Help students find their class locations, check their weekly schedule, and get reminders for upcoming deadlines.
Navigate Campus: Provide directions to buildings and offices on a digital map.
Natural Conversations: Understand and respond to questions in a natural, conversational way, not just with rigid commands.
How we built it
Frontend: A clean and responsive web interface built with React.js and CSS to ensure it works seamlessly on both desktop and mobile.
Backend: We used Node.js with Express.js to create a robust server handling user requests and business logic.
Chatbot & AI: The core intelligence is powered by a large language model (like GPT-4 or an open-source alternative like Llama 3) via an API. We used LangChain to structure our prompts and manage the conversation flow effectively.
Data: We created a custom knowledge base by scraping and structuring official university data (websites, PDFs, handbooks) and stored it in a vector database (Pinecone or ChromaDB) for fast and accurate semantic search.
Integration: All components are dockerized and deployed on a cloud platform like AWS or Google Cloud.
Challenges we ran into
Data Scraping & Structuring: Consolidating inconsistent and unstructured data from various university websites into a clean, reliable knowledge base was a significant hurdle.
Hallucination Control: Preventing the AI from "hallucinating" or inventing incorrect information about university policies was critical. We implemented a robust Retrieval-Augmented Generation (RAG) pipeline to ground its answers strictly in our provided data.
Context Management: Designing the system to maintain context throughout a conversation (e.g., remembering a student was asking about "the science building" when they follow up with "where is the cafeteria there?") required careful state management.
Integration Complexity: Getting the frontend, backend, AI model, and database to communicate seamlessly and with low latency was a complex task.
Accomplishments that we're proud of
Creating a Truly Useful Tool: We built a functional prototype that can accurately answer a wide range of real student questions, validating our core concept.
Mastering RAG: Successfully implementing a RAG system that drastically reduced AI inaccuracies and made the bot trustworthy.
Clean User Experience: Designing an intuitive and friendly chat interface that feels helpful, not robotic.
Teamwork: Collaborating effectively under time constraints, with team members specializing in frontend, backend, and AI to create a cohesive final product.
What we learned
The importance of data quality: An AI is only as good as the data it's trained on. Data cleaning and preparation was 80% of the work.
Prompt Engineering is key: Small changes in how we prompt the AI model can lead to massive differences in response quality and accuracy.
The power of modern AI tools: Frameworks like LangChain and vector databases abstract away immense complexity, allowing small teams to build powerful AI applications.
User-Centric Design: We learned to anticipate the diverse ways students phrase their questions and design the system to be as flexible as possible.
What's next for a unibot
Personalization: Integrate with student portals (with secure authentication) to provide personalized information like grades, GPA, and specific financial aid status.
Proactive Notifications: Send push notifications for deadline reminders, class cancellations, or when a hold is placed on a student's account.
Campus Marketplace Integration: Allow students to ask about textbook swaps, club events, or roommate finder services.
Multimodal Input: Enable the bot to understand image uploads (e.g., a photo of a campus map) or provide voice-based interaction.
Pilot Program: Partner with our university's student union or IT department to run a beta test with a small group of real students and gather feedback for improvement.
Log in or sign up for Devpost to join the conversation.