Hong Kong PolyU’s Top AI Scientist Yang Hongxia Eyes The ‘Last Mile’ Of Generative AI
Yang Hongxia, one of Hong Kong Polytechnic University’s leading artificial intelligence researchers, is setting her sights on what she calls the last mile of generative AI. Instead of relying on a handful of major technology companies to supply large language models, she wants hospitals, factories, financial firms and other industries to build AI systems tailored to their own needs.
Yang believes that the next breakthrough in the global AI race will not come solely from building bigger models, but from making these systems usable and affordable for every organisation. This vision is the foundation of her new start up, InfiX.ai, which she launched after years developing advanced models at ByteDance and Alibaba’s Damo Academy.
Moving Beyond One Size Fits All AI
In her recent conversation with the South China Morning Post, Yang explained that many enterprises operate in highly specialized fields where general purpose AI tools struggle to provide accurate or safe results. A hospital, for instance, handles medical terminology, patient records and regulatory constraints that are vastly different from the needs of a logistics company or a retail brand.
Her solution is to build domain specific LLMs, allowing organisations to train, refine and control their own AI systems. Instead of depending entirely on the models produced by big firms, these customised models would act as intelligent assistants deeply familiar with industry specific knowledge.
A Future Where AI Is Truly Decentralised
InfiX.ai’s technical mission is clear. The company aims to develop open frameworks and tools that help businesses fine tune models efficiently, using far fewer resources than the cost of training full scale LLMs.
According to the start up’s pages on GitHub and Hugging Face, the long term goal is to support what it calls decentralised generative AI. This refers to a broad ecosystem in which AI development is not concentrated among a few global giants, but shared among hospitals, universities, manufacturers and small businesses that maintain their own specialised intelligence.
Yang sees this as a foundation for a more balanced digital future, where access to powerful AI does not depend on financial power or geographic location.
Making AI Practical For Real World Use
Yang’s approach also focuses on practicality. Many companies hesitate to adopt generative AI because of concerns over data security, training cost or unpredictability in model outputs. InfiX.ai plans to address these issues directly, enabling organisations to maintain full control of their data and customise AI behaviour based on strict operational needs.
Her team is also working on methods that reduce the hardware requirements for model training. This could allow companies to run advanced AI systems on smaller machines or cloud infrastructure without overwhelming budgets, making adoption far more accessible.
Bringing Innovation Back To The User
Yang often describes AI not as a replacement for human expertise, but as a bridge that helps professionals work smarter. In medicine, domain specific LLMs could help doctors analyse reports more quickly. In logistics, they could predict transportation risks or delays. In education, personalised AI models could assist teachers in managing large classrooms more efficiently.
She believes that by giving organisations their own tools, innovation no longer needs to come solely from top down development by tech giants. Instead, real breakthroughs will emerge from those who know their industries best.
The New Phase Of Hong Kong’s AI Ambition
Yang’s work also reflects Hong Kong’s growing ambition to become a major regional AI hub. PolyU, where she teaches, has invested heavily in machine learning laboratories and deep tech research. The launch of InfiX.ai adds to the city’s expanding ecosystem of start ups focused on AI applications, robotics and digital transformation.
Her experience at two of China’s most influential tech companies brings credibility and momentum to this effort, while her focus on decentralisation fits within global discussions on fair access to artificial intelligence.
