Gen-AI Education Platform

RAG is an AI architecture that combines a retrieval system with a language model. When a user asks a question, the system first retrieves relevant content from the knowledge base — PDFs, webpages, or video transcripts — then passes that retrieved context to the language model to generate an accurate, grounded answer. This prevents the hallucination problem of pure LLMs because the model answers based on retrieved source content rather than its parametric memory alone.

The chatbot gave users instant, accurate answers to questions that previously required contacting support — drawn directly from the platform's own PDFs, help pages, and video content. Instead of waiting for a human response, users got relevant answers through the natural language interface in seconds. As the knowledge base was indexed, the chatbot handled the full tail of repetitive queries consuming most of the support team's time.

Cohere is an enterprise AI platform providing large language model APIs optimised for business applications — with strengths in retrieval-augmented generation, semantic search, and multi-language support. It was chosen for its strong RAG performance, predictable API costs at scale, and enterprise data privacy guarantees — important for a platform handling user-generated content across different formats.

Yes. GYSP's multi-format RAG architecture uses separate ingestion pipelines for each content type — PDF parsing, web crawling and page extraction, and YouTube transcript processing — all feeding into a unified vector store. When a question is asked, the retrieval layer searches across all content types simultaneously, synthesising an answer from whichever format contains the most relevant information for that specific query.