The Urban Data Futures initiative is the cornerstone of AURIN’s commitment to innovation and cross-sector collaboration; it equips emerging researchers with the power to apply data-driven solutions for the urban challenges of Australia.
Empowering the next generation of researchers
The Urban Data Futures initiative at AURIN equips data science students with data literacy, skills and practical experience needed to engage with real-world urban and regional challenges. AURIN enables this by providing technical advisory support, real-world use-cases and hands-on learning opportunities.
The program places a significant focus on advanced urban data management, analysis and visualisation, such as GenAI and machine learning, geospatial problem-solving, Python programming and both relational and non-relational databases.
The goal is to blend academic discipline with practical experience to prepare these future leaders of urban data for impactful roles in data-driven decision-making.
This year, AURIN hosted 4 groups of final-year Master of Data Science students from the University of Melbourne’s Faculty of Science to collaborate on their capstone projects. A group of three AURIN data scientists, Dr Masoud Rahimi, Dr Hao Chen and Dr Bradley Geig led the initiatives, overseeing projects’ progress and mentoring the 19 participating students.
Let’s have a look at the projects in more detail:
GenAI for smarter data quality assurance
Can generative AI (GenAI) improve data quality assurance beyond traditional rule-based checks? Our initial review showed that existing tools, like the FAIRs FAIR Automated FAIRness Data Assessment Tool (F-UJI), are limited as they cannot reason qualitatively or check geospatial data, often producing incomplete or misleading results.
To address this, our students built a GenAI-powered system that combines automated data and metadata checks with AI reasoning. Their framework uses large language models to interpret data, guided feedback loops to refine results, and structured tracking to keep the process transparent. This approach allows for smarter, cross-checking of different data quality principles and produces clear, traceable justifications, capabilities missing in traditional rule-based tools.
The project speeds up data validation, reduces manual effort, and improves the reliability of complex urban datasets. It also gave students practical experience in AI-driven data quality work, building both their technical and domain knowledge.
You can see a demonstration of their project here.
Urban research collaboration network and expert recommendation system
Can knowledge graphs help decision-makers find the right urban research experts? In Australia, urban research is spread across universities, government, and industry, yet finding the right experts is often difficult. Information about researchers is scattered, inconsistent, and hard to search, making collaboration slow and inefficient. To this end, our students developed a prototype that enables users to find collaborators based on expertise and niches with the use of natural language queries. The prototype gets a query from a decision maker, digests the request, identifies and recommends relevant experts to the problem and explains why they were recommended. This is enabled through effective data acquisition and wrangling, a Neo4j knowledge graph, a recommendation system, and a simple yet effective natural-language interface.
The graph maps authors, works, institutions, and topics from OpenAlex, cleaned and deduplicated, reflecting a snapshot from September 2025. The recommendation system supports different query types: one path finds specific researchers or profiles using a GraphSAGE model trained on author collaborations, while another handles open-ended text queries with a hybrid scoring method. The natural-language interface allows users to search using plain text.
While still a proof of concept, limited to Australian urban researchers and a single data source, this project showcases that graph-based expert discovery can make connecting decision-makers and researchers faster, easier, and more transparent.
View a demonstration of their project here.
AURIN’s Urban Grant Finder: Enhancing access to funding for urban research
Can AI help urban researchers find the right funding faster?
Finding suitable grants can be frustrating for urban researchers, planners, and policymakers because funding landscapes are complex and fragmented. A group of students tackled this challenge by building the Urban Grant Finder, an AI-driven tool that matches funding opportunities to a researcher’s profile and interests.
The PoC collects data from APIs and websites, uses natural language processing and large language models to understand queries, and applies a graph-based recommendation engine to suggest grants. Grants, programs, institutions, and eligibility rules are modelled as interconnected entities in a Neo4j graph, allowing the tool to provide transparent, explainable recommendations. Users can search with plain language and receive alerts for relevant opportunities, improving efficiency and collaboration across academia, industry, and government.
While still a proof-of-concept, the project shows that combining AI, knowledge graphs, and recommendation systems can simplify funding searches, increase success rates, and open new avenues for urban research innovation.
View a demonstration of their project here.
Data format and validation registry
Can data validation be automated to ensure standards compliance and improve data integrity and interoperability? Modern datasets are increasingly complex, coming in varied formats, sizes, and sources, which makes manual validation both challenging and error-prone. Automated data validation offers a solution, ensuring standards compliance, improving data integrity, and supporting interoperability across systems like AURIN.
To tackle this, students developed a Data Format and Validation Registry that combines modular validation rules with a user-friendly interface. The system automatically detects dataset formats and types, applying relevant rules to identify geometric, topological, and semantic errors. Compact, dataset-specific checkers maintain consistency, while a lightweight Validation Controller handles data loading, normalisation, and rule execution. Validation results and metadata are stored in a registry, ensuring full traceability and reproducibility.
The dashboard enables users to upload datasets, toggle recommended validation rules, and visualise results via interactive maps. This approach creates a scalable, collaborative, and fully traceable workflow for managing data quality, making automated validation accessible to both technical and non-technical users.
View a demonstration of their project here.
Learn more and take part
The Urban Data Futures initiative shows AURIN’s commitment to impactful research and the development of the next generation of data leaders.
Dedicated mentoring, access to our national digital research infrastructure, and collaborative networks contributed to ongoing positive change by AURIN in Australia’s urban and regional research ecosystem.
To find out more about our work and how to get involved, contact us at support@aurin.org.au