PolyU academic-led startup wins BEYOND Best of Innovation Awards, garnering international industry recognition for translating frontier AI research into enterprise solutions
The Hong Kong Polytechnic University (PolyU) has once again gained international recognition for its research excellence in the field of artificial intelligence (AI). InfiX.ai, an AI infrastructure startup founded by Prof. YANG Hongxia, Executive Director of the PolyU Academy for Artificial Intelligence (PAAI), Vobile Group Professor in Generative Artificial Intelligence, Associate Dean (Global Engagement) of the Faculty of Computer and Mathematical Sciences, and Chair Professor of Generative AI at the Department of Computing, received the BEYOND Best of Innovation Awards 2026 for InfiCube, its integrated large-model training and inference appliance. This accomplishment demonstrates the University’s strong capabilities in frontier AI research and knowledge transfer.
BEYOND Expo 2026, held in Macao from 27 to 30 May, is one of Asia’s most influential platforms for technology innovation and industry ecosystems. The BEYOND Best of Innovation Awards represent the highest honour within the BEYOND Innovation Awards, selected from among award winners to recognise breakthrough technologies and industry-leading innovative companies and products. InfiCube received this prestigious accolade alongside products from renowned technology companies from around the world, marking international industry acknowledgement of the PolyU research team’s commercialisation capabilities in enterprise-grade AI infrastructure, integrated large-model training and inference technologies.
On the final day of the Expo, Prof. Yang participated in a panel discussion themed “The AI-Native Enterprise: Architecting B2B Intelligence for the Global Stage”. She shared insights on how enterprises in the AI-native era can build their own domain models while ensuring data security and cost-effectiveness, and discussed how low-bit training, private deployment and next-generation AI infrastructure support reliable deployment of large models in business-to-business scenarios.
Prof. Zijian ZHENG, Vice President (Knowledge Transfer) of PolyU, commended the award-winning startup team, stating, “The distinguished performance of this PolyU academic-led startup fully demonstrates the University’s comprehensive strengths in fundamental AI research, technological innovation and knowledge transfer, and underscores our commitment to advancing the commercialisation of research outcomes and to fostering innovation and entrepreneurship. Through our holistic flagship startup ecosystem PolyVentures, the University provides all-round support to research teams and entrepreneurs, facilitating the translation of research into market-ready solutions. The University will continue to deepen collaboration across the government, industry, academia and research sectors, translating further cutting-edge technologies into practical applications and incubating more excellent startups, contributing to Hong Kong’s development as an international innovation and technology hub.”
The award-winning InfiCube is InfiX.ai’s integrated large-model training and inference appliance designed for enterprise AI scenarios, and represents a landmark milestone in the commercialisation of frontier AI technologies. The product was jointly developed by InfiX.ai and its ecosystem partners. InfiX.ai leads the development of large-model training and inference algorithms, while working with high-performance computing platforms and underlying system support to form an integrated hardware-software AI infrastructure covering model training, fine-tuning, inference deployment, permission management and continuous optimisation.
In terms of data security, InfiCube’s entire infrastructure supports the completion of training, fine-tuning and inference within localised deployment environments, allowing sensitive enterprise data to remain inside private environments rather than being uploaded to external systems. The system can also incorporate permission management and audit records, making it suitable for industry scenarios with higher data security and compliance management requirements. For continuous optimisation, InfiCube’s built-in machine learning operations capabilities enable ongoing industry data integration, incremental training, feedback-based fine-tuning and model iteration. This allows AI to move beyond one-time deployment tools and become a continuously validated and optimised capability system embedded in enterprise workflows.
Prof. Yang is an AI scientist with extensive experience in both academic research and industrial practice. Her research encompasses large models, model training and inference, model fusion, AI infrastructure, and intelligent industry applications. Leading the PAAI team, Prof. Yang continues to achieve breakthroughs in Generative AI (GenAI), proposing the Collaborative Generative AI (Co-GenAI) paradigm that shifts AI training from traditional centralised, monolithic approaches to decentralised ones. This approach significantly lowers training resource requirements, protects data privacy and removes resource barriers, enabling more global research institutions to participate in AI research and development. Her team has also developed multiple world-leading technological advancements, including an end-to-end FP8 low-bit training solution and InfiFusion model fusion technology.
Prof. Yang’s research projects have received support from the Government of the Hong Kong Special Administrative Region (HKSAR) of the People’s Republic of China and industry. Her project “Reallm: World-leading Enterprise GenAI Infrastructure Solution” has been funded under the Research, Academic and Industry Sectors One-plus (RAISe+) Scheme of the Innovation and Technology Commission of the HKSAR Government. Her pioneering project “Collaborative Generative AI (Co-GenAI)” has received funding from the Theme-based Research Scheme 2025/26 under the Research Grants Council. In addition, InfiX.ai has obtained the highest funding ratio under Cyberport’s Artificial Intelligence Subsidy Scheme. All these achievements demonstrate substantial industry recognition of her technological innovation and industrial value in Co-GenAI and AI-native infrastructure.
In knowledge transfer, Prof. Yang advances frontier AI research from academic achievements to industrial applications. Building on research foundations in GenAI, model fusion, low-cost training and enterprise-grade AI infrastructure, her team has translated PolyU’s research outcomes through InfiX.ai and its core product InfiXspeed into a domain AI development and deployment platform serving real-world enterprise scenarios. The platform supports data processing, model training, evaluation, inference deployment and business integration, helping enterprises build controllable, deployable and continuously optimisable domain expert AI systems.
PolyU is a pioneer in driving innovation and entrepreneurship. Through its holistic flagship startup ecosystem PolyVentures, the University has nurtured over 600 active startups to date, including two listed companies, three unicorns and 26 ponies. The University actively responds to the HKSAR’s innovation and technology development direction, connecting research with industry needs and advancing the translation and adoption of AI and emerging technologies. PolyU will continue to refine its PolyVentures, deepen collaboration across the government, industry, academia and research sectors, and nurture further high-quality research outcomes, contributing to the innovation and technology development of Hong Kong and our Nation.
***END***
Achievements
Knowledge Transfer and Entrepreneurship Office