Dr Sebastian Haan1, Mrs. Kirsten Jackson1
1The University of Sydney, Newtown, Australia
Biography:
Seb is a machine learning and AI specialist at the Sydney Informatics Hub, University of Sydney, with expertise in probabilistic models, computer vision, geospatial ML, and AI agent networks. With a background in particle and astrophysics, he has held research positions at Caltech, CSIRO, and the Max-Planck-Institute. An active open-source contributor and published author in major scientific journals, Seb is passionate about developing innovative AI applications that accelerate research and create real-world impact. He brings extensive data-driven research experience to tackle emerging challenges in the rapidly evolving AI landscape.
Abstract:
Research impact evaluation is crucial for funding and policy decisions, yet traditional assessment methods remain confined to bibliometric and funding metrics, often overlooking real world benefits. Evidence scattered across disconnected sources makes comprehensive evaluation both time-consuming and incomplete. ResearchPulse automates the Research Impact Assessment Framework (RIAF) by combining Large Language Models with intelligent evidence discovery to generate impact assessments and use-cases within minutes.
We developed ResearchPulse featuring a Context Agent that uses Large Language Models to iteratively identify knowledge gaps and proactively search for missing evidence across heterogeneous data sources — from academic databases to policy documents, web content, and structured data catalogues. The system embeds RIAF criteria directly into its reasoning process, using semantic understanding to connect research outputs with their real-world impacts.
The platform synthesises findings from multiple sources into coherent narratives, linking research activities to measurable impacts whilst maintaining accurate source referencing. The system provides explainable AI decisions at each assessment stage, ensuring transparency in the evaluation process. Feedback integration enables researchers to refine assessments and customise indicators. Trials at the University of Sydney and Monash University demonstrate that ResearchPulse reduces assessment time from weeks to minutes whilst maintaining evidence-based rigour through automated fact validation.
As research funding becomes increasingly competitive and impact-driven, this web-based platform transforms how institutions evaluate, document, and communicate research outcomes. Field-specific templates enable adaptation across academic disciplines. This presentation will showcase the technical architecture, demonstrate practical implementation, and outline our roadmap for making the platform accessible to all Australian research organisations.