AMBS researcher releases open collection of financial AI models

26 Nov 2025

Congratulations to Dr Eghbal Rahimikia, who has released a large open collection of financial AI models built using billions of data points and backed by UK government investment in AI infrastructure.

Dr Eghbal Rahimikia, Assistant Professor of Financial Technology in AMBS, has released a substantial open collection of financial AI models developed through a large international collaboration. As the project lead and a contributor to the ‘AI at Manchester’ initiative, Dr Rahimikia oversaw the work from its initial design through data preparation, model development, extensive evaluation, and the release of the collection. He collaborated closely with colleagues at University College London and Shanghai University, bringing together expertise in AI, mathematics, and finance.

The research supporting this work, presented in the study ‘Re(Visiting) Time Series Foundation Models in Finance’, draws on billions of financial observations covering 94 global markets. Using one of the most extensive financial datasets assembled for this purpose, the research examines how various types of AI models, particularly the recently introduced time series foundation models (TSFMs), perform when forecasting financial returns.

The open collection is intended to support researchers, students, and industry practitioners who want to explore responsible uses of AI in financial contexts and to help develop new approaches to understanding market behaviour.

To support this work, the project secured two rounds of access to the UK's new national AI computing infrastructure, with allocations of resources amounting to approximately £0.5m. This access made it possible to draw on the national AI Research Resource, including the Isambard-AI supercomputer. This programme aims to expand national AI capacity by investing in high-performance computing facilities dedicated to artificial intelligence research.

The release of this open collection reflects an ongoing commitment to transparency, reproducibility, and collaboration in AI research. All models are accessible through the FinText.ai portal and additional tools, code, and documentation will be released to help other researchers build on the work. By making these resources openly available, the project aims to support wider research communities, contribute to education and training in AI for finance, and encourage the development of new methods that help explain and interpret financial markets.