Research projects in Information Technology

Classical propositional logic (CPL) captures our basic understanding of the linguistic connectives “and”, “or” and “not”. It also provides a very good basis for digital circuits. But it does not account for more sophisticated linguistic notions such as “always”, “possibly”, “believed” or “knows”. Philosophers therefore invented many different non-classical logics which extend CPL with further operators for these notions.

Classical propositional logic (CPL) captures our basic understanding of the linguistic connectives “and”, “or” and “not”. It also provides a very good basis for digital circuits. But it does not account for more sophisticated linguistic notions such as “always”, “possibly”, “believed” or “knows”. Philosophers therefore invented many different non-classical logics which extend CPL with further operators for these notions.

Neuro-symbolic AI combines the strengths of neural and symbolic methods to efficiently learn and reason over models of the world. Typically, many of the assurances that can be provided by such systems are limited to the learning errors due to the neural component. In this Ph.D. project, you will be exploring the use of Lipschitz Continuous Neural Networks to learn Lipschitz-bounded neural models of the world to create a Neuro-symbolic AI that can provide assurance guarantees.

Neuro-symbolic AI combines the strengths of neural and symbolic methods to efficiently learn and reason over models of the world. Typically, many of the assurances that can be provided by such systems are limited to the learning errors due to the neural component. In this Ph.D. project, you will be exploring the use of Physics-Informed Neural Networks to encode the symbolic knowledge into the learned neural models of the world to create a Neuro-symbolic AI that can provide assurance guarantees.

Planning is the reasoning side of acting in Artificial Intelligence. Planning automates the selection and the organisation of actions to reach desired states of the world as best as possible. For many real-world planning problems however, it is difficult to obtain the full model of the world that captures its complex dynamics. Fortunately, the unknown parts model can be accurately approximated as neuro-symbolic (deep) neural networks which then can be used in planning.

This project will result in a much fuller understanding of the state of the Birrarung than is currently possible, as well as qualitative and quantitative results to model different interventions and their effect on swimmability.

The project will build tools and techniques to understand and decide on effective interventions to improve the Birrarung’s swimmability.

In many branches of science (e.g., Artificial Intelligence, Engineering etc.), the modelling of the problem is done through the use of functions (e.g., f(x) = y). On a very high-level, we can think of Machine Learning as the problem of approximating function f from the pair of measurements (x,y), and Optimization as the problem of finding the value of input x that maximizes the output y given function f.

In NeuroDistSys (NDS): Optimized Distributed Training and Inference on Large-Scale Distributed Systems, we aim to design and implement cutting-edge techniques to optimize the training and inference of Machine Learning (ML) models across large-scale distributed systems. Leveraging advanced AI and distributed computing strategies, this project focuses on deploying ML models on real-world distributed infrastructures, improving system performance, scalability, and efficiency by optimizing resource usage (e.g., GPUs, CPUs, energy consumption).

Public transportation is vital for sustainable urban mobility, yet challenges like inefficient first- and last-mile connectivity, and over-reliance on private cars hinder its effectiveness. Autonomous vehicles (AVs) offer transformative potential by enabling diverse, on-demand mobility solutions tailored to specific trip needs, enhancing connectivity, and reducing emissions. However, current research often overlooks the complexities of mixed-vehicle environments, and the development of optimal deployment, routing, and charging strategies.

Automating code generation, SQL query formulation, and data preprocessing pipelines is a crucial step toward intelligent and efficient software development. This project aims to leverage large language models (LLMs) to address these challenges by developing a comprehensive framework that seamlessly integrates LLM capabilities for generating accurate and optimised code, constructing complex SQL queries, and automating data transformations.