Honours and Masters project

Modern NLP applications increasingly process text carrying sensitive personal information, including clinical conversations, legal correspondence, customer support transcripts, and social media posts. Sharing such text with third-party models, annotators, or downstream pipelines remains constrained by data protection legislation (e.g., GDPR, the EU AI Act) and growing user expectations around transparency.

Modern large multimodal models (LMMs) and omni-modal models process not just text but vision, audio, and speech, opening new application surfaces and, with them, new safety risks. Established safety pipelines, including RLHF, safety classifiers such as Llama Guard, and red-teaming protocols, were largely developed for text-only models and translate poorly to the multimodal setting. Three gaps are now well documented.

Legal professionals routinely spend significant time on case analysis tasks that follow well-defined reasoning protocols, of which the Issue-Rule-Application-Conclusion (IRAC) method is the most widely taught and practiced framework across common-law jurisdictions.

Team‑based projects are widely used across computing education to support the development of technical competence alongside collaboration and professional skills. While students engage extensively in teamwork during these projects, educators often face challenges in seeing and responding to teamwork processes as they unfold, which can constrain opportunities to provide timely, process‑focused feedback beyond final project outcomes.

This project investigates how user and developer behaviour can be modelled as latent states underlying observable software-engineering and requirements-engineering artefacts, and how recovering these states can deliver actionable insight to practitioners — for example, early signals of requirement instability, indicators of stakeholder misalignment, or behavioural predictors of defect-prone modules.

Background and Motivation

The project aims to evaluate an immersive virtual reality system for visual exploration of global data. Visual exploration of maps often requires a contextual understanding at multiple scales and locations. Multiview map layouts, which present a hierarchy of multiple views to reveal detail at various scales and locations, have been shown to support better performance than traditional single-view exploration on desktop displays. We created a virtual reality system, named immersive multiview maps, that allows for visual exploration of global data across geographical and temporal scales. 

Online articles, including news and government reports, hold critical significance in communicating environmental issues (e.g., Black Summer bushfires, water security, or renewable energy). Although charts and photographs on a 2D screen can communicate facts, they struggle to cultivate the deeper engagement and empathy that comes from direct presence in the affected environments.

Two converging technological shifts create a new opportunity.

In EdgeVLMOpt (EVO): Optimizing Vision-Language Models for Resource-Constrained Edge Devices, we aim to develop efficient and scalable techniques to enable the deployment of advanced vision-language models (VLMs) on edge hardware. While VLMs have demonstrated strong capabilities in multimodal reasoning and understanding, their high computational and memory demands pose significant challenges for real-time, on-device applications.

In EdgeFusionAI (EFAI): Real-Time Multi-Sensor Multi-Modal Intelligence on Edge Devices, we aim to design and develop efficient techniques for fusing heterogeneous sensory data, including vision, LiDAR, radar, and other modalities, to enable robust and real-time decision-making on resource-constrained edge platforms. This project focuses on building intelligent systems capable of integrating diverse data sources while addressing the challenges of limited computation, memory, and energy availability at the edge.