DrShahadat Uddin

Project Opportunities

Title: Fair Machine Learning in Predictive Modelling

Summary of opportunity:

This research project will use large-scale datasets to explore and understand the fairness of different machine-learning algorithms deployed in various contexts for predictive modelling. 

Opportunity synopsis:

With the rapid advances in high-tech computing in recent years, AI-based systems have increasingly been used across numerous sensitive environments to make essential, life-changing decisions. Therefore, we must ensure these decisions are fair and bias-free. They do not show discriminatory behaviour or bias towards specific groups or populations. The current literature shows that different real-world AI-based systems tend to make unfair and biased decisions, mainly due to the data used, the algorithms, and the users. In addition to causing false alarms in the underlying context, such unfair outcomes will inherently create greater bias in training future algorithms. This research project will explore different facets of fair machine learning and its impact across various domains, and develop approaches to detect bias to help us avoid unfair AI-based decisions. 

Successful candidates must:

• Have a Bachelor's degree (Honours or 1st class honours equivalent) or a Master's degree
• Have a strong back in machine learning and statistical modelling
• Have a strong background in data analytics and Python (or similar)

How to Apply: To apply, please email shahadat.uddin@sydney.edu.au the following:

• CV
• Transcript

--------------------------------------------------

Project Opportunities

Title: Predictive disease modelling using Big data analytics and Complex network methods

Summary of opportunity:

This research project aims to mine the administrative claims dataset to develop predictive models for individual chronic diseases and for the comorbidity of multiple chronic diseases. For this purpose, it will use techniques and methods from Big Data analytics and Complex networks.

Opportunity synopsis:

Chronic diseases and their subsequent tendency to develop comorbid conditions are a significant health concern worldwide. Apart from causing premature mortality and reduced quality of life, these diseases exert a significant ongoing treatment cost, which impacts national economies. Prevention of chronic diseases has been a major challenge since patients, in many cases, remain unaware of the progression until they are admitted to a hospital. On the other hand, the administrative claim dataset is the largest single source of over-time utilisation data for different healthcare services. This dataset has already attracted significant attention for research, as it covers a wide range of medical services across a broad geographic area. Using big data analytics and complex network methods, this research project will mine administrative claim data to develop predictive models of the natural history of different chronic diseases and their tendency to develop comorbid conditions.

The project requires:

• A good background in Data Analytics
• Expert in a programming language (Python)

How to apply: To apply, please email shahadat.uddin@sydney.edu.au the following:

• CV
• Academic transcripts

--------------------------------------------------

Project Opportunities

Title: Prompt Engineering to Improve Fairness in Large Language Models

Summary of opportunity:

This research project aims to explore and develop prompt engineering techniques to improve fairness (i.e., mitigate biases) in large language models.

Research Areas:

Fair AI; Large Language Model

Opportunity synopsis:

Although the world is witnessing an unprecedented application of Large Language Model (LLM)- based generative AI systems across various domains, these systems often produce biased or unfair outcomes. Adequate and appropriate prompt engineering is crucial for improving fairness or reducing biases in LLMs. Such a practical approach to LLMs involves understanding the nuances of language and context to create prompts that encourage more equitable responses across different demographic groups. By carefully designing and refining the input prompts used to guide these models, it is possible to mitigate biased outputs that often reflect societal prejudices. This project aims to develop prompt engineering techniques to improve fairness (i.e., reduce biases) in large language models.

The successful candidate must:

• Have experience in AI, Large Language Model and Generative AI
• Strong background in data analytics and Python (or similar)

How to apply: To apply, please email shahadat.uddin@sydney.edu.au the following:

• CV
• Academic transcripts

--------------------------------------------------

Project Opportunities

Title: Data Bias and Model Fairness in Unsupervised Machine Learning

Summary of opportunity:

This project investigates the impact of data bias on the fairness and reliability of unsupervised machine learning algorithms. While fairness has been extensively studied in supervised contexts, unsupervised models, such as clustering and dimensionality reduction, remain vulnerable to latent biases in input data that can propagate through analyses and decision-making pipelines. The research aims to develop theoretical frameworks and practical methodologies for detecting, quantifying, and mitigating bias in unsupervised learning, focusing on promoting equitable outcomes across diverse applications, including healthcare, education, and social data analytics.

Keywords:

Fair AI, Data Bias, Unsupervised Machine Learning

The successful candidate must:

• Have a strong background in Data Analysis, Machine Learning and Python (or similar)

How to apply: To apply, please email shahadat.uddin@sydney.edu.au the following:

• CV
• Academic transcripts