PPMS - Submit Project Preferences

CITS5553 - Data Science Capstone Project

Term: Semester 2 - 2026

Unit Coordinator: Wei Liu

Are you Perth based? If so, do you have data, be it in PDFs, CSVs, texts, images, audios or videos? We have multiple teams of 5-6 capable Data Science students at UWA, available for Second Semester (Aug-Oct), 2025, to tackle your data-related problems of interest. They can apply their solid skills in data engineering, visualisation, machine learning, natural language processing and GenAI techniques to help prototyping the extraction of business values out of your data assets. You are warmly invited to register you interests by filling in this short survey.

Feel free to contact Wei Liu ([email protected]) directly if you are data-rich and ready to share your real-world data analytics needs to help preparing our soon-to-be data scientists/engineers/analysts/AI-Architects to join our future workforce.

Group Information

Only use this form for a team's preferences

Group Name *

Group Number (if assigned)

Representative Name *

Representative Email *

Reason for Project Selection * Why do these projects interest your group? This information helps coordinators understand your preferences and make better project assignments.

Your Selected Projects

Only use this form for a team's preferences

How to Rank Projects:

1. Select projects below by checking the box. 2. Drag and drop to reorder your selections.

Your selected projects will appear here. Drag to reorder.

* At least one project must be selected before submitting

Available Projects

8 Projects

GSWA HyLogger (1 of 3): Data Crunch

This project, offered by the Geological Survey of Western Australia (GSWA), is the first of three related proposals; this one asks a UWA Data Science Capstone team to characterise GSWA’s …

This project, offered by the Geological Survey of Western Australia (GSWA), is the first of three related proposals; this one asks a UWA Data Science Capstone team to characterise GSWA’s HyLogger hyperspectral drill core dataset, accessed through the National Virtual Core Library (NVCL). The dataset spans thousands of drill holes and hundreds of kilometres of core, but has not yet been analysed as an integrated whole, so the team would start by profiling its structure, scale, and statistical properties, including how key mineralogical variables vary by depth, geology, and geography. Building on this foundation, the team would apply techniques such as clustering, dimensionality reduction, and anomaly detection to surface patterns in the spectral data, and explore fusing HyLogger outputs with complementary datasets such as whole-rock geochemistry and petrophysical measurements to reveal lithological or alteration signatures. Deliverables could include interactive data profiles, statistical summaries, predictive models of mineralogical abundance, or reproducible analytical pipelines that GSWA staff can adopt directly. GSWA geoscientists are available to provide domain context and curated data access, with no prior geoscience background required of the student team. The completed project will also be eligible for submission to GSWA’s open HyLogger hackathon event, where it will be judged by a panel from GSWA, academia, and industry alongside entries on the other two themes, with cash prizes awarded to outstanding submissions.

Client: GILLFEATHER-CLARK, Tasman

Location: 100 Plain St, Perth WA 6004

Multiple Teams

IP: Creative Commons (open source)

Video‑Derived Behavioural Modelling

Our laboratory investigates circadian rhythms using animal models and has collected hundreds of hours of video data from sheep over recent years. Despite rapid advances in computer vision, we have identified a key methodological gap: there is currently no integrated, end‑to‑end pipeline that links raw video processing to behavioural pattern discovery, behavioural annotation, and ultimately circadian rhythm analysis. In preliminary work, we piloted a pose‑based behavioural analysis approach to begin addressing this challenge. An overview of our prototype pipeline is available here: https://github.com/user-attachments/assets/49c39c32-7cbc-4f84-a175-7c457dd068d0, along with an example output video here: https://github.com/user-attachments/assets/654c28a5-69b6-4507-94a9-ffb0c72b199e. This project offers students the opportunity to design and implement a scalable, end‑to‑end analysis pipeline for video actigraphy. Students will have access to substantial computational resources via the UWA High Performance Computing Cluster (Kaya), enabling exploration of high‑throughput methods. There is considerable scope for creativity, including experimenting with state‑of‑the‑art computer vision techniques (e.g., pose estimation and foreground detection) and machine learning methods for unsupervised pattern discovery and temporal modelling of behaviour.

Client: Harrison Smith

Location: UWA Crawley Campus

IP: Creative Commons (open source)

AI-Powered Service Knowledge Management System for Research Infrastructure

Research infrastructure facilities such as CMCA at UWA generate substantial amounts of documentation, including service reports, maintenance records, instrument logs, troubleshooting notes, operating procedures, incident reports, and vendor documentation. Much …

GSWA HyLogger (2 of 3): Industry Adoption

This project, offered by the Geological Survey of Western Australia (GSWA), is the second of three related proposals; this one asks a UWA Data Science Capstone team to improve industry …

This project, offered by the Geological Survey of Western Australia (GSWA), is the second of three related proposals; this one asks a UWA Data Science Capstone team to improve industry uptake of GSWA’s HyLogger hyperspectral drill core dataset, made accessible via the National Virtual Core Library (NVCL). Despite its richness, the dataset sees uneven use in the exploration industry due to accessibility barriers and unfamiliarity with spectral data formats, so the team would focus on building a translation layer between raw spectral data and practical, industry-ready insight. Particular emphasis is on near-mine applications such as grade control, alteration mapping, and ore-waste discrimination, alongside related domains including resource evaluation, geometallurgical modelling, and environmental baseline assessment, where HyLogger data could complement or partly substitute for costly laboratory analysis. The team could develop a decision-support dashboard, an automated classification pipeline, an integration connector for existing industry workflows, or a case study demonstrating value in a specific mining context. The goal is a compelling, practitioner-facing demonstrator, not just a proof of concept, supported throughout by GSWA geoscience experts. The completed project will also be eligible for submission to GSWA’s open HyLogger hackathon event, where it will be judged alongside entries on the other two themes, with cash prizes awarded to outstanding submissions.

Client: GILLFEATHER-CLARK, Tasman

Location: 100 Plain St, Perth WA 6004

Multiple Teams

IP: Creative Commons (open source)

GSWA HyLogger (3 of 3): Visualisation

This project, offered by the Geological Survey of Western Australia (GSWA), is the third of three related proposals; this one asks a UWA Data Science Capstone team to make GSWA’s …

This project, offered by the Geological Survey of Western Australia (GSWA), is the third of three related proposals; this one asks a UWA Data Science Capstone team to make GSWA’s HyLogger hyperspectral drill core data, hosted via the National Virtual Core Library (NVCL), easier to navigate and understand. Spectral data is inherently high-dimensional and hard to interpret with specialist tools alone, so the team would design visualisation and interaction approaches aimed at geoscientists, students, regulators, and non-specialist audiences alike. Possible directions include loading HyLogger data into GIS platforms such as QGIS for spatial, map-based exploration of mineralogical variation, building 3D visualisations of spectral profiles along drill traces within a geological model, or developing web-based interactive viewers, animated cross-sections, comparative drill hole displays, or uncertainty visualisations. The emphasis throughout is on clarity, usability, and honest representation of data quality. GSWA domain experts would support the team in understanding the dataset and validating that outputs are demonstrable to a non-specialist audience. The completed project will also be eligible for submission to GSWA’s open HyLogger hackathon event, where it will be judged alongside entries on the other two themes, with cash prizes awarded to outstanding submissions.

Client: GILLFEATHER-CLARK, Tasman

Location: 100 Plain St, Perth WA 6004

Multiple Teams

IP: Creative Commons (open source)

AI-Powered Virtual Patient Avatar for Communication Skills Training and Assessment in Podiatry Education

Effective communication is a critical component of podiatry practice, yet providing consistent, authentic, and scalable opportunities for students to develop these skills can be challenging. This project aims to develop an AI-powered virtual patient avatar that can engage in realistic, two-way conversations with students during simulated podiatry consultations. The avatar would have a visual presence, verbal communication capabilities, and the ability to respond dynamically to student questions and behaviours, creating an interactive learning experience that closely mirrors real-world clinical encounters. The virtual patient could be designed to represent a wide range of clinical and social scenarios, including culturally safe care, disability, mental health presentations, challenging consultations, health literacy barriers, and complex patient interactions. By scripting core patient characteristics and responses, educators can ensure all students are exposed to equivalent learning opportunities, supporting consistency and fairness across the cohort. The platform could also allow educators to develop and modify patient cases, enabling a diverse library of simulated consultations tailored to specific learning outcomes. A key feature of the project would be the integration of assessment and feedback tools. Using embedded rubrics and artificial intelligence, the system could identify critical moments within the consultation, evaluate communication behaviours against predefined criteria, and provide structured feedback to both students and educators. This would create a safe environment for students to practise adaptive communication strategies, receive immediate feedback, and demonstrate competence before engaging with real patients. Students involved in developing this project would gain valuable experience in conversational AI, avatar development, natural language processing, user experience design, educational technology, and healthcare simulation.

Client: Renee Silvester

Location: Perth

NDA Required

Multiple Teams

IP: Open to further discussion about best IP options.

Smart Assessment Platform for Rural and Remote Clinical Education Using Point-of-View Smart Glasses

As smart glasses become increasingly integrated into health professional education, there is an opportunity to develop a purpose-built platform that supports student learning, assessment, and reflection during clinical placements. This project aims to create a web-based software solution, similar to GoReact, that enables students to upload point-of-view video recordings captured through smart glasses while undertaking clinical activities. The platform would be designed to support learning in rural and remote placement settings, where direct supervision and access to educators may be limited. The platform would include features such as video review for self-reflection, timestamped tagging of key clinical moments, and integrated assessment rubrics that can be customised by educators. Additional functionality could include artificial intelligence tools to assist with rubric-based feedback, identify moments aligned with learning outcomes, and generate prompts that encourage deeper student reflection. These features would enhance the efficiency and consistency of assessment while supporting student engagement with reflective practice. A unique aspect of the project would be the ability to capture and synchronise multiple perspectives of the same clinical encounter, such as the student’s point of view through smart glasses, an external observer or assessor view, and potentially a patient perspective. This multi-view approach would provide richer educational insights, support assessment moderation, and create new opportunities for research into clinical education and feedback practices. Students involved in this project would gain experience in software development, user experience design, artificial intelligence integration, video processing, and educational technology innovation within the healthcare sector.

Client: Renee Silvester

Location: Perth

NDA Required

Multiple Teams

IP: Open to further discussion about best IP options.

Pump Predictive Maintenance

Analyze pump data and make predictions using machine learning about whether the pump will require a maintenance intervention in the near future (20-30 minutes). This approach is inspired by: https://arxiv.org/pdf/2508.19974 This project has industrial interest from the Perth office of Programmed a facilities management and maintenance services company. Maintenance activities across industry are currently primarily classified as either reactive maintenance, meaning fixing or replacing something after it has failed or proactive maintenance, meaning that machines and infrastructure are replaced or rebuilt on a schedule with a frequency meant to stay ahead of failure. The problem with reactive maintenance is that running machines or infrastructure to failure results in unexpected shutdowns for maintenance. Proactive maintenance on the other hand may lead to "over-servicing" in which machines and infrastructure are replaced too far in advance, leading to increased costs and because this maintenance is "schedule based" there can still be unexpected failures. The promise of predictive maintenance is that by use of continuous monitoring and machine learning (or deep learning) failure can be predicted before it happens and maintenance interventions can take place based on the actual condition of the machine or infrastructure rather than on a predefined schedule.

Client: Pete Whittaker

Location: UWA School of Engineering, Crawley Campus. Building 224, room 220.

Multiple Teams

IP: Creative Commons (open source)

You can rearrange the projects any time before submitting.

Ranking Guidelines

Consider Team Skills

Choose projects that align with your team's technical abilities and interests.

Evaluate Requirements

Check if project has NDA or IP requirements that your team can accommodate.

Rank Strategically

Consider ranking a mix of popular and less competitive projects to increase chances.

View Projects or Submit Project Preferences

CITS5553 - Data Science Capstone Project

Group Information

Only use this form for a team's preferences

Your Selected Projects

Only use this form for a team's preferences

How to Rank Projects:

Available Projects

GSWA HyLogger (1 of 3): Data Crunch

Video‑Derived Behavioural Modelling

AI-Powered Service Knowledge Management System for Research Infrastructure

GSWA HyLogger (2 of 3): Industry Adoption

GSWA HyLogger (3 of 3): Visualisation

AI-Powered Virtual Patient Avatar for Communication Skills Training and Assessment in Podiatry Education

Smart Assessment Platform for Rural and Remote Clinical Education Using Point-of-View Smart Glasses

Pump Predictive Maintenance

Ranking Guidelines

Consider Team Skills

Evaluate Requirements

Rank Strategically