AIML Grand Discovery Challenge

The Australian Institute for Machine Learning (AIML) Grand Discovery Challenge is a machine learning competition designed to solve real-world problems using applied AI, engaging the talents and skills of undergraduate and postgraduate students at Adelaide University. 

The inaugural AIML Grand Discovery Challenge is inviting interdisciplinary student teams to apply machine learning to real datasets provided by researchers within the University, industry or government, to solve specific data-driven problems. Students will work in teams to develop predictive models, collaborating closely with leading researchers and experts in machine learning. The challenge offers a chance to build their portfolio, showcase their skills, and compete for recognition and potential future research opportunities. Prizes to be announced. 

We are currently seeking sponsors to contribute to the competition’s prize pool. In addition, Adelaide University researchers as well as sponsors are invited to provide datasets that they would like to have explored using machine learning techniques. This offers a unique opportunity to gain insights into your own data while supporting the development of emerging talent. 

Key dates for 2025 

• Sponsorship and dataset deadline: Friday, 8 August 2025
• Competition launch: Monday, 8 September 2025
• Team registration deadline: Friday, 19 September 2025 
• Competition closes: Friday, 28 November 2025 
• Winners announced: 8 December 2025 

Competition Overview

• The competition will provide a problem statement and a dataset, typically split into training data (which includes labels) and test data (which does not). 

• Participants develop and train machine learning models using the training data, then make predictions on the test set. 

• Submissions are evaluated against unseen ground-truth data, and results are displayed on a leaderboard (with both public and private versions to ensure fair evaluation and limit overfitting). 

• The evaluation process will assess model performance (i.e. accuracy, F1-score, RMSE), code clarity and documentation, reproducibility of results, and practical relevance to real world applications. 

• Participants must also provide a brief technical report or Jupyter notebook with source code explaining methodology, preprocessing, and model design. 

• There are usually limits on the number of daily submissions to prevent brute-force approaches. 

• When the competition ends, final rankings are determined using the private leaderboard, and prizes are awarded to the top performers. 

• Top teams will be invited to present their work to representatives at AIML. 

The use of external data or public benchmarks is not permitted for this competition. Each team may only make one daily submission of original work. Competition organisers reserve the right to disqualify teams for any misconduct. 

Forming a Team

• The AIML Grand Discovery Challenge is open to students at Adelaide University (including the University of Adelaide and the University of South Australia) enrolled in an undergraduate or postgraduate degree. 

• There is a maximum of 3 members per team. Interdisciplinary participation is welcome and encouraged. 

Data Specifications

In line with the research priorities of the future Adelaide University, we are seeking datasets aligned with the following themes: 

• Creative and cultural 

• Food, agriculture and wine 

• Sustainable green transition 

• Defence and national security 

• Personal and societal health 

For more information on these themes, please visit: adelaideuni.edu.au/research  

Adelaide University researchers or sponsors who are providing data for the competition must ensure their datasets are: 

• Under 1 GB, with ~1000 samples. 

• Not commercially sensitive. 

And, with limited assistance from AIML researchers: 

• Can be split into a training and test set. 

• Are annotated with ground-truth labels for evaluation. 

Resources

The AIML Grand Discovery Challenge is modelled on Kaggle competitions, online competitions where participants build and submit machine learning models to solve specific data-driven problems using provided datasets. Visit the Kaggle website for more information on this format.