Curtin HDR Chatbot

Brief

We partnered with Curtin University to develop a secure, offline chatbot that supports international HDR students in finding accurate information quickly. Powered by Qwen 3 (a publicly available LLM) and a custom RAG pipeline, the system delivers multilingual, reference-linked answers sourced directly from official Curtin documents and webpages. Hosted entirely on Curtin infrastructure, it provides fast, reliable guidance without relying on external APIs or cloud services.

We highly commend the AI team at AI Advancements. What stood out most was their clear and responsive communication, as well as their genuine dedication to customer satisfaction.

Sonny Pham, Associate Professor at Curtin University

Problem

HDR students face a complex web of policies, forms, guidelines and support resources scattered across multiple Curtin University websites and documents. Information existed, but students often struggled to locate what they needed, especially when English was not their first language. The research team had previously used a basic GPT prototype, but it required manually uploaded documents, had no verification layer, and could not guarantee the accuracy or trustworthiness of responses. They needed a system that could:

• Run fully offline for privacy and cost certainty
• Pull information from a large quantity of PDFs and webpages
• Always reference an official source rather than generate unsupported answers
• Support multilingual interactions
• Be easy to update as new content becomes available This project formed the foundation of a broader research initiative into how AI can improve the HDR student experience.

Solution

We built a complete locally hosted chatbot ecosystem powered by Qwen 3 and a tailored Retrieval-Augmented Generation pipeline. The system was designed for maximum accuracy, transparency and security. Key features include:

• Offline LLM deployment: Qwen 3 running entirely on Curtin-owned hardware, ensuring no data ever leaves the university’s environment.
• RAG-based information retrieval: Content from PDFs and scraped HTML pages is indexed, allowing the chatbot to respond only with verifiable, reference-backed answers.
• Website scraper + document extractor: The system automatically processes Curtin webpages and PDF resources, extracting both text and the original links. Students can click through to confirm the information themselves.
• Multilingual support: Qwen 3 enables students to ask questions in their preferred language while still receiving accurate references.
• Accuracy-first behaviour: If no sufficiently relevant source is found, the chatbot does not guess. Instead, it redirects students to official support channels.
• Internal-only access: The entire website and LLM stack are hosted on Curtin infrastructure, providing security, privacy and long-term cost control. The solution provides a clear and maintainable foundation for the research team to build upon as more data is collected.

Result

The chatbot now gives HDR students a much faster, clearer path to the information they need. Early feedback shows that students are finding answers more easily and spending less time searching across scattered pages and documents. For the research team, the system provides a secure, extensible platform that can continue to grow as more content is added. By running the entire solution offline, Curtin benefits from stronger data control, predictable costs and a robust technical foundation for future AI projects.