Honours and Masters project

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.

Bundle recommendation systems enhance user experience and increase sales by recommending a set of items as a bundle rather than individual items [1].   The understanding of items in bundles is that they should be complementary some how.  In this project, we will explore the relationship between causal reasoning on items purchased and bundles.  Causal reasoning could be used to infer if two purchases are related, and, moreover, language language models can be used to assess the plausability of this, to help create an argument using the Bradford-Hill criteria.  Some recent research has…

AI is starting to be used for reviewing in international AI conferences.  For instance, http://paperreview.ai as announced by Andrew Ng as an “Agentic Reviewer” trained on ICLR 2025 reviews, and Google is field testing an AI-based Paper Assistant Tool for authors submtting to NeurIPS 2026.  Significant research is going into developing these tools, and some good test data exists collected from previous conferences.  In this project we will work with the VinUniversity AI group who is developing a benchmarking framework for evaluating AI reviewing.  Research papers usually contain important…

It is said that ChatGPT is ‘not trying to be right, it’s just trying to be plausible.’  While the LLM community talk about hallucination, it is a complex phenomenon and a product of their construction.  The training and theory of LLMs has no notion of truth, they generate text with no critical evaluation of content or sources.  Many text sources are opinions, some may have subtle or not so subtle propaganda, some reflecting misinformed views, and LLMs reproduce this mess.  Moreover, the training and theory of LLMs has no notion of epistemic uncertainty, they have no sense of uncertainty. …

This is a Winter Student Research Internship ONLY not an honours or minor thesis project at this time.

Please apply here if you are interested in the role before the deadline: 

https://www.monash.edu/study/fees-scholarships/scholarships/summer-winter

___

This is a Winter Student Research Internship. Please apply here: https://www.monash.edu/study/fees-scholarships/scholarships/summer-winter

Advances in artificial intelligence (AI) are reshaping work across computing-related professions. Although AI systems can perform some technical and cognitive tasks, they continue to rely on human judgement, responsibility, and contextual understanding. Identifying where human contribution remains critical is therefore a substantive question for disciplines concerned with socio-technical systems.

Pharmacists handle thousands of medications daily, requiring constant verification of prescriptions, allergy checks, and inventory management across multiple digital and physical systems. This fragmented workflow leads to frequent context switching, high cognitive load, and increased risk of workflow inefficiencies and medication errors. 

This project addresses the language ignoring problem in embodied AI, where robots learn visual shortcuts instead of following instructions. Building on our preprint establishing the relationship between planning width (instruction granularity) and learning difficulty, you will develop WALR—a reward design framework that adapts to instruction complexity. WALR scales language grounding rewards based on instruction granularity (coarse vs.