Yang R1, Guo E1, Ahmed M1, Sun Y1
1National Computational Infrastructure, Australian National University, Canberra, Australia
Biography:
Rui works as a Senior HPC Specialist at the National Computational Infrastructure (NCI), where he leads the Software and Data Modernisation team. The team provides dedicated support to multiple domain research communities by maintaining specialised computing environments, optimising performance for scientific models, and supporting collaborative projects. Rui led initiatives in deploying and evaluating state-of-the-art AI/ML models on HPC systems, maintaining ML-ready datasets and co-develop AI workflows with research groups to meet domain-specific needs.
Abstract:
The rapid emergence of artificial intelligence (AI) is transforming scientific research across disciplines. At the same time, researchers face growing challenges in accessing large-scale datasets, managing complex software environments, deploying AI models, and efficiently using advanced computing resources. To address these challenges, the National Computational Infrastructure (NCI) is developing an integrated AI-enabled research platform that brings together scientific data, AI models, software environments, and supercomputing resources within a unified ecosystem.
The platform supports the end-to-end AI-for-Science workflow, from data discovery and visualisation through to model training, fine-tuning, inference, benchmarking, and analysis. Large collections of climate, weather, environmental, and geoscience datasets are exposed through standardised intake catalogues and interactive interfaces, enabling researchers to easily discover and access AI-ready data. Reproducible containerised environments provide consistent access to scientific software, machine learning frameworks, and research workflows.
The platform also hosts a growing collection of state-of-the-art AI models for weather forecasting, climate prediction, downscaling, and Earth-system applications. Through a common inference and evaluation framework, researchers can deploy, compare, and benchmark models against local datasets, accelerating both scientific experimentation and operational applications. Built on NCI’s national supercomputing infrastructure, the platform enables scalable training and inference across CPU and GPU resources while shielding researchers from much of the underlying infrastructure complexity.
Currently encompassing more than 80 data collections and 50 software environments, the platform is enabling hundreds of users to adopt AI methods through integrated data services, reusable workflows, and accessible analysis tools. This integration reduces barriers to entry for AI-driven research and shortens the time from data access to model evaluation.
Positioned within a broader international shift towards AI-enabled research infrastructures, this work demonstrates how the integration of AI-ready data services, model ecosystems, inference capabilities, and supercomputing resources can accelerate reproducible, data-intensive scientific discovery. It also presents architectural design choices and operational lessons learned from building and deploying this capability.
