Learn with us

🌟 Inspiration

The idea for Learn With Us was born from a common struggle we all faced as students — endless PDFs, scattered notes, and the constant stress of preparing for both exams and placements. Traditional study tools focused only on memorization, not on understanding or career readiness. We wanted to create something smarter — an AI-driven platform that could not only summarize, explain, and quiz you on your study material, but also help you develop real-world skills alongside academic knowledge. That’s how Learn Flow came to life — a multi-agent study companion that learns with you and for you.

🧩 How We Built It

We designed Learn With Us as a multi-agent AI system using LangGraph and LangChain, where each agent specializes in a task — one for document-based Q&A, another for skill generation, another for study planning, and one for quizzes.

The backend is built with FastAPI, orchestrating the AI agents and connecting to the vector database (Qdrant Cloud) for semantic search and retrieval.

The frontend uses Streamlit, offering a clean and interactive interface for uploading PDFs, chatting with the AI, generating skills, or taking quizzes.

Google Gemini API powers the intelligence — enabling context-aware reasoning, planning, and creativity — while Google Embeddings handle document vectorization for retrieval-augmented generation (RAG).

This modular architecture allows seamless collaboration between agents, giving Learn Flow its adaptive, multi-purpose personality.

💡 What We Learned

Building Learn With us taught us more than just technology — it taught us design thinking. We learned how to coordinate multiple AI agents effectively within a single workflow. We explored RAG pipelines, vector databases, and prompt engineering for precise question answering. We gained experience in frontend–backend integration, environment configuration, and deploying APIs that interact with external AI models. Most importantly, we learned that good AI design isn’t about complexity — it’s about clarity, reliability, and real usefulness for end users.

⚙️ Challenges We Faced Every stage brought unique challenges: Vector database integration: Managing embeddings and indexing large PDFs efficiently in Qdrant took careful optimization. Agent coordination: Ensuring agents didn’t overlap or contradict each other required refining the LangGraph flow logic. API management: Handling Gemini API rate limits and environment configuration across backend and frontend. Frontend responsiveness: Making Streamlit dynamic enough to handle multi-agent outputs without freezing or lagging. Despite these challenges, the process strengthened our understanding of multi-agent systems, contextual reasoning, and the future of AI-driven education.

🚀 The Outcome Learn Flow now serves as an all-in-one AI study companion that helps students revise smarter, learn faster, and prepare better — both for academics and for their careers. It’s not just an assistant; it’s the bridge between learning and becoming.