PAIGE

AI CHATBOT FOR TUM STUDENT SERVICES — BUILT WITH MICROSOFT GERMANY

ROLE

FRONTEND & AI ENGINEER

TEAM

TUM + MICROSOFT GERMANY

DURATION

DEC 2024 — PRESENT

TOOLS

FRONTEND, POSTGRESQL, WEAVIATE, LANGCHAIN

OVERVIEW

pAIge is an AI-powered chatbot serving TUM students with admissions, thesis, and general inquiry support. Developed in collaboration with Microsoft Germany, it uses retrieval-augmented generation to provide accurate, contextual answers from TUM's knowledge base.

TECHNICAL APPROACH

The system uses LangChain for orchestrating the RAG pipeline. Weaviate serves as the vector database, storing embeddings of TUM documentation, FAQs, and policy documents. PostgreSQL handles conversation history and user sessions. The frontend provides a clean chat interface with citation support.

WON 1ST PLACE AT THE MICROSOFT X TUM HACKATHON. THEN BUILT THE REAL THING.

KEY CONTRIBUTIONS

RAG PIPELINE — Designed the retrieval system with hybrid search (keyword + semantic) for high-accuracy responses

FRONTEND — Built the chat UI with streaming responses, citation rendering, and conversation management

DATA INGESTION — Created pipelines to parse and embed TUM documentation into Weaviate

EVALUATION — Built testing framework to measure response accuracy against ground truth Q&A pairs

← PREVIOUS

MAKKR.AI

NEXT →

ADHD ROBOTICS