Lingnan University wins 16 awards at Silicon Valley International Invention Festival 2026 - Tops Hong Kong universities in total and gold medals

Awards

Gold Medal;

IFIA Best Invention Award;

Thailand Award for the Best International Invention and Innovation

Project Title

GaitGPT: Evidence-Grounded AI for Bidirectional Clinical Gait Translation

Winning Lingnan Faculty and Staff

·        Prof S. Joe Qin, President and Wai Kee Kau Chair Professor of Data Science

·        Prof Yu Lisha, Assistant Professor of Teaching of the Division of Artificial Intelligence of the School of Data Science and Programme Director of the Doctor of Artificial Intelligence Studies Programme

·        Mr Zhang Weijian, Graduate of the Division of Artificial Intelligence of the School of Data Science

·        Mr Hou Haochen, Graduate of the Division of Artificial Intelligence of the School of Data Science

Project Description

GaitGPT is an AI-powered gait analysis system that bridges the longstanding gap between qualitative clinical descriptions and quantitative gait indicators. By connecting three core pillars, users seeking health insights, professionals providing expertise, and multi-sensor data streams, through a patented constraint-enforced architecture. GaitGPT makes professional gait assessment and management explainable, reliable, and evidence-based for everyone, everywhere.

Key Features

·            Evidence-Threshold Gating: A three-gate architecture that sequentially enforces evidence sufficiency, measurement feasibility, and biomechanical coherence before any output is generated, preventing unsupported or clinically inconsistent recommendations.

·            Bidirectional Translation Engine: Translates in both directions between clinical semantic phrases and quantitative gait indicators, grounded in a structured evidence graph rather than unconstrained language generation.

·         Dual-View Reporting with Privacy-First Design: Delivers both professional-facing and data-facing outputs with full audit trails, supported by a self-hosted architecture that ensures data sovereignty across lab, clinic, and home settings.

Awards

Gold Medal

Project Title

Face Fortress: Adaptive All-In-One AI for Digital ID Safety

Winning Lingnan Faculty and Staff

·        Prof Sam Kwong Tak-wu, Associate Vice-President (Strategic Research), Dean of the School of Graduate Studies and J.K. Lee Chair Professor of Computational Intelligence

·        Dr Guo Haifeng, Postdoctoral Fellow of the Division of Artificial Intelligence of the School of Data Science

Project Description

The innovation of this system lies in the seamless integration of deepfake detection and quality screening into a unified verification workflow. It first evaluates key quality metrics such as image clarity and lighting automatically, then uses advanced algorithms to identify whether synthetic faces exist, and finally performs facial recognition. It is particularly well-suited for identity verification or maintaining public safety applications.

Key Features

·        Dual-Gate Protection: By combining deepfake detection and quality screening into a single workflow, the system eliminates the cumbersome integration steps of running multiple systems, enhancing operational efficiency and reliability.

·        Image Filtering: It automatically assesses multiple quality indicators of uploaded facial images, including clarity, lighting, and angle, and actively filters blurred or substandard images to ensure the accuracy of subsequent verification processes.

·        Critical Scenario Suitability: Tailored for applications with stringent accuracy requirements such as bank account opening or immigration security checks, the system delivers reliable, robust results.

Awards

Gold Medal

Project Title

Water-Retentive Hydrogel Smart Windows: Cooling Without Power

Winning Lingnan Faculty and Staff

·        Prof Chen Xi, Dean and Chair Professor of Interdisciplinary Studies of the Wu Jieh Yee School of Interdisciplinary Studies

·        Prof Ke Yujie, Assistant Professor of the Wu Jieh Yee School of Interdisciplinary Studies

·        Zhao Yu, PhD student at the Wu Jieh Yee School of Interdisciplinary Studies

Project Description

This is a mass-producible hydrogel smart window that regulates indoor temperatures autonomously without electricity. Combining a water-retentive thermochromic hydrogel with a low-emissivity film, the technology filters solar heat radiation to achieve indoor cooling, and reduces energy consumption by up to 34 per cent.

Key Features

·        Autonomous Regulation: Relying entirely on the inherent physical properties of the materials, the system senses temperature shifts autonomously and adjusts dynamically in real time, achieving heat insulation in summer and thermal retention in winter.

·        Significant Energy Saving and Carbon Reduction: Compared with traditional architectural glass, this invention cuts indoor cooling energy consumption by up to 34 per cent, promoting environmental sustainability.

Awards

Gold Medal

Project Title

PCG.ME: Infinite Game Co-Creation Platform with Behaviour-Adaptive AI

Winning Lingnan Faculty and Staff

·        Prof Liu Jialin, Associate Professor of the Division of Industrial Data Science of the School of Data Science and Programme Director of the Master of Science in Smart City Technologies and Applications Programme

·        Ms Li Yuchen, Assistant Research Officer of the Division of Industrial Data Science of the School of Data Science

Project Description

PCG.ME is a human-AI, collaborative, web-based platform for game creation. Users simply explain their requirements in everyday language or adjust intuitive controls, and the platform uses AI to automatically generate, evaluate, and refine levels and rules. Creators maintain full supervision and editing capabilities throughout, ensuring that game design is no longer constrained by technical barriers.

Key Features

·        Multiple AI Testing: Combining AI playtesting, automated fine-tuning to balance difficulty and engagement, and leveraging large language models, generative models, reinforcement learning, and multi-objective optimisation  to translate creative ideas into substantive game content, the platform delivers experiences that are both playable and original.

·        Behaviour-Adaptive Adjustment: The system observes and analyses player interactions in real time, dynamically generating novel subsequent game content to allow the game world to evolve alongside the player.

·        Creator-Led: Emphasising designer involvement at every stage, all AI-generated candidate options can be inspected, compared, and refined by creators, ensuring that creative control remains firmly in human hands.

·        Player-led: Enhancing player engagement, not only are the game contents adaptively generated based on player experience, but the platform also allows the player to create his/her own game.

Awards

Gold Medal

Project Title

FlexiBot: Make Every Parking Spot a Charging Spot

Winning Lingnan Faculty and Staff

·        Prof Tang Xiaopeng, Assistant Professor (Presidential Early Career Scholar) of the Division of Science

·        Mr Wang Fangyu, Senior Research Assistant of the Division of Science

Project Description

As the adoption of electric vehicles accelerates, the scarcity of parking bays and the underuse of overnight charging points have become daily frustrations for drivers. FlexiBot is an intelligent robotic solution designed to provide wireless charging for electric vehicles in any parking space. Able to navigate freely around car parks and slide underneath vehicles to dock automatically, it is compatible with all electric vehicle brands and delivers a seamless park-and-charge experience. The system also features intelligent scheduling, automatically avoiding peak electricity hours to reduce grid load and save costs.

Key Features

·        Flexible Autonomous Charging: The robot navigates automatically to the vehicle's location to deliver a charge, allowing a single unit to service several vehicles. This eliminates the expensive infrastructure modification costs associated with traditional car park trenching and cable laying.

·        High Compatibility: Compatible with all electric vehicle brands due to dedicated adapters, the system operates safely and conveniently, supported by a dedicated mobile app.

·        Intelligent Scheduling: By automatically arranging charging times and making optimal use of overnight off-peak periods, the system increases operational efficiency by approximately 300 per cent while optimising energy use.

Awards

Gold Medal

Project Title

MedCenterDet: AI System for Accurate Lesion Detection in 3D Medical Images

Winning Lingnan Faculty and Staff

·            Prof Pan Fei, Assistant Professor of the Wu Jieh Yee School of Interdisciplinary Studies

·            Mr Zeng Qiang, Research Assistant of the Wu Jieh Yee School of Interdisciplinary Studies

Project Description

MedCenterDet is an intelligent analysis algorithm for 3D medical images such as CT and MRI scans. Unlike traditional methods that can only roughly outline lesion areas or organs, this algorithm acts like a precise "smart radar" that can directly lock onto the centre of the target.

Key Features

·        Smart locking and flexible adaptation: This algorithm does not rely on any fixed templates or assumptions. Instead, it directly finds the "core coordinates" of each organ or lesion area and automatically adjusts the judgment range according to its actual size. Regardless of the target's size or shape, it achieves precise localisation.

·        Powerful background filtering and auto-correction: Medical scan images often contain complex background information such as bones and blood vessels, which can cause visual confusion. This algorithm effectively eliminates noise, focuses on the actual lesion areas, and can automatically detect and correct minor positional deviations.

·        Efficient and time-saving, reducing doctors' burden: The algorithm runs efficiently with low computational demands, greatly reducing the time doctors spend on manual annotation, accelerating the image analysis process, and taking a crucial step towards future fully automated diagnosis.

Awards

Gold Medal

Project Title

StorySketcher: A Multi-Agent AI System for Child-Led Co-Creative Storytelling

Winning Lingnan Faculty and Staff

·        Prof Xu Xian, Assistant Professor of the Department of Digital Arts and Creative Industries

Project Description

This platform uses AI-guided questioning to turn children's drawings into unique stories. Once a story is complete, it automatically generates an animated short film complete with lively narration within minutes, turning imagination into a visual work that can be shared and enjoyed, stimulating children's narrative creativity, logical thinking, and verbal ability, and making creation both effortless and enjoyable.

Key Features

·        Inquiry-Based Inspiration: Acting as a collaborative partner, the system translates visuals into a short film with subtitles while asking questions to inspire children to extend their storylines. This helps them to construct a complete narrative framework step-by-step, yet ensures the direction of the story remains entirely up to the child.

·        Co-Creation and Multilingual Learning: Designed around a drawing board interface that is simple and straightforward, the platform can be operated independently  by children, but remains intuitive enough for parental guidance and companionship, allowing families to enjoy the fun of collaborative creation together. The system also supports multilingual output in Chinese, English, French, and other languages, turning the creative process into a cross-lingual learning experience.

Awards

Gold Medal

Project Title

IDEAL-Gen.AI - Instructional Design Enhanced Active Learning through Generative Artificial Intelligence

Winning Lingnan Faculty and Staff

Dr Ronnie H. Shroff, Principal Project Fellow of the Centre for Innovative Teaching and Learning

Project Description

This platform uses generative AI and large language models to help educators create personalised learning activities within seconds. Available for free in both Chinese and English editions, it enhances instructional design with greater efficiency and creativity, reducing lesson preparation time by 80 to 90 per cent and allowing for greater focus on student engagement and personalised instruction

At the QS Reimagine Education Awards and Conference 2025, the platform won the Bronze Award in the Best Use of AI category.

Key Features

·        Tailored Teaching: Powered by advanced AI algorithms, teachers without formal pedagogical training can generate high-quality, subject-specific lesson plans with a few clicks. This makes personalised teaching and learning effortless, and fully supports diverse educational needs from kindergarten to higher education.

·        Intelligent Filtering: Featuring pioneering automated prompt-generation technology, the platform allows teachers to filter precisely by subject, student proficiency, and class size, ensuring that the automatically generated content aligns perfectly with real-world classroom requirements.

·        Forward-Looking Assessment: Having generated over 14,000 AI-driven outputs as of July 2026, the platform will soon introduce AI-ready assessment tools that will help educators transform traditional assessments into AI-resilient and ethically AI-inclusive assessment tasks.

Awards

Silver Medal

Project Title

GeoAI Mosquito Risk Prediction Platform — Kill, Monitor, Identify & Forecast

Winning Lingnan Faculty and Staff

·        Prof Paulina Wong Pui-yun, Head and Associate Professor (Presidential Early Career Scholar) of the Division of Science and Programme Director of the Master of Science in Sustainability and Environmental Analytics Programme

·        Prof Tang Xiaopeng, Assistant Professor (Presidential Early Career Scholar) of the Division of Science

·        Mr Wang Fangyu, Senior Research Assistant of the Division of Science

Project Description

This platform is Hong Kong's first mosquito forecasting system to combine geospatial artificial intelligence (GeoAI), artificial intelligence of things (AIoT), and geospatial science and multimodal signal fusion technique. By deploying solar-powered smart mosquito traps and meteorological stations, it collects and identifies real-time data on mosquito species and weather conditions, and generates a Mosquito Risk Index and heat maps, forecasting mosquito abundance levels  three days in advance. The technology has already been implemented in several Hong Kong Housing Society estates, enabling proactive mosquito prevention.

Key Features

·        Dynamic Predictive Prevention: The smart traps eliminate mosquitoes and record catch numbers and species data, while meteorological stations capture real-time weather information such as temperature, relative humidity, and rainfall. Together, they generate a Mosquito Risk Index and risk maps, enabling frontline property management staff to deploy resources earlier in high-risk areas.

·        Intelligent Mosquito Species Identification: Leveraging advanced AI and multimodal signal fusion technology, the system automatically identifies different mosquito species, including vectors for dengue fever and chikungunya, providing robust scientific evidence to support more targeted, species-specific support for prevention and control policies.

·        Solar-Powered Operation: Requiring only a few hours of indirect sunlight to sustain a full week of operation, the traps use built-in lithium batteries designed to withstand all weather conditions.

·        Sustainable Green Solution: A 100 per cent solar-powered system equipped with battery storage, engineered for seamless integration into urban ecosystems. It delivers resilience against extreme weather conditions while driving significant operational cost savings

Awards

Silver Medal

Project Title

Generative AI Assessment System (GAAS): A Game-Changer for Timely Student Feedback

Winning Lingnan Faculty and Staff

·        Prof Frankie Lam King-sun, Director of the Centre for Innovative Teaching and Learning

Project Description

The platform provides practical guidance, pedagogical guidance, interactive examples, ethical guidelines, curated AI tools, and a knowledge hub to help educators redesign assessment practices. Key features include the Generative Artificial Intelligence Assessment System (GAAS) that streamlines the process and improves the consistency of feedback provided to students and supports formative and student-centred approaches to assessment, drawing upon advances in artificial intelligence (AI) and natural language processing (NLP). 

GAAS is a “game-changer” for assessment in the AI era.  It reduces the manual effort involved in producing high quality assessments across courses. It provides rapid, consistent and continuous feedback on student assignments while minimising human errors or biases.  More importantly, GAAS identifies strengths, weaknesses, learning outcome attainment, areas for enhancement, and supports scalable implementation across courses.

Key Features

·        Automated Real-Time Feedback: Combining AI technology, the system provides students with immediate, targeted learning feedback to improve learning effectiveness and experiences in assignments that include capstone projects and service-learning initiatives. It assists teaching teams simultaneously in optimising assessment mechanisms to improve overall learning outcomes.

·        Privacy Protection: Safeguarding student privacy is paramount. All cases submitted to the system for evaluation are automatically de-identified to omit students’ names and identification numbers, and all data collected are permanently deleted after use.

·        Driving Personalised Learning: The system identifies the strengths and weaknesses of student assignments accurately, and makes specific, actionable suggestions for improvement. Students can instantly view their progress, and teachers gain a comprehensive view of their learning trajectories and can promote student growth through educational assessment.

Awards

Silver Medal

Project Title

BuildGuard: Blockchain Oracle AI for Instant Secure Compliance in Offsite Manufacturing

Winning Lingnan Faculty and Staff

·        Prof Mike Wu, Assistant Professor of the Division of Industrial Data Science of the School of Data Science

Project Description

This invention combines secure, tamper-proof blockchain technology with intelligent, AI-driven auditing to provide direct compliance verification for cross-border offsite manufacturing, including Modular Integrated Construction (MiC). The system transmits diverse data types securely into a highly encrypted blockchain ledger for real-time analysis, guaranteeing that records remain tamper-proof. It also transforms traditional paper-based manual inspections into instantaneous digital workflows, addressing issues of low data credibility, human error, and difficult retrospective tracing. This supports the large-scale development of offsite manufacturing, ensuring compliance with the latest regulatory requirements and safeguarding engineering quality.

Key Features

·        Real-Time Compliance Auditing: By training AI models to analyse multi-modal production floor data in real time, the system performs expert-level compliance checks autonomously.

·        Blockchain Tamper-Proof Mechanism: By integrating blockchain oracle technology, inspection data are securely transmitted to a blockchain ledger, guaranteeing authentic and immutable audit results.

·        Streamlined Inspection Workflows: By converting time-consuming manual compliance checks into instantaneous digital processes, the system facilitates offsite manufacturing efficiency and high-capacity production.

Awards

Silver Medal

Project Title

IDShield: A Privacy Firewall Against Behavioural Re-identification in AI Dialogues

Winning Lingnan Faculty and Staff

·        Prof Shen Jiaxing, Assistant Professor of the Division of Artificial Intelligence of the School of Data Science and Programme Director of the Master of Science in Data Science Programme

·        Wang Wenxuan, PhD student in the Division of Artificial Intelligence of the School of Data Science

·        Liu Zirui, PhD student in the Division of Artificial Intelligence of the School of Data Science

·        Kou Haoxuan, PhD student in the Division of Artificial Intelligence of the School of Data Science

Project Description

Removing names, account details, or IP addresses does not necessarily make an AI conversation anonymous. Across repeated interactions, users may leave behind distinctive “behavioural fingerprints” through their topics of interest, writing style, interaction patterns, and personality-related expressions. These subtle signals can allow separate anonymous conversations to be linked to the same individual, creating a largely overlooked privacy risk.

IDShield addresses this vulnerability through a working, locally operated privacy gateway positioned between a user application and an external AI service. Before a prompt leaves the user’s or organisation’s trusted environment, IDShield detects personally identifiable information and identity-revealing content, pseudonymises protected values, and rewrites potentially linkable content and writing-style signals. It then performs safety checks for semantic fidelity, formatting, protected fields, and residual privacy risk. Information required to complete the task can subsequently be restored locally after the external AI service returns its response. The system does not require the underlying language model to be modified or retrained. Its current research prototype operates through a Python API and command-line interface and can function as middleware for enterprise model APIs and conversational AI systems.

Key Features

·        Evidence-based behavioural fingerprint assessment: IDShield is evaluated against a dedicated attack model that measures whether anonymous conversations can be linked across sessions through content, writing style, interaction patterns, and personality-related signals. This provides a measurable technical indicator of privacy risk rather than relying solely on the removal of obvious identifiers.

·        Proactive, locally controlled de-identification: Unlike conventional privacy measures that focus mainly on masking accounts, names, or IP addresses, IDShield processes prompts before they are transmitted to an external AI service. It combines local detection, reversible pseudonymisation, and controlled rewriting to reduce both explicit and behavioural identity signals.

·        Privacy protection with built-in utility safeguards: IDShield verifies semantic preservation, formatting, protected fields, and residual risk before releasing a transformed prompt. Unsafe transformations can trigger a fallback procedure or be blocked, while task-critical protected values can be restored locally after the AI response is received.

·        Model-independent integration: IDShield can operate as a privacy middleware layer without modifying or retraining the target language model. The prototype can be integrated with enterprise APIs, including OpenAI, Azure OpenAI, and other model providers, as well as organisation-operated conversational AI systems.

·        Technical evidence for privacy and AI-governance reviews: IDShield does not certify or guarantee legal compliance. Instead, it produces technical evidence that may support organisational reviews under the GDPR and, where applicable, the EU AI Act. Its risk measurements, pseudonymisation records, safety checks, residual-risk assessments, and privacy-preserving audit records may contribute to data-minimisation, privacy-by-design, security, impact-assessment, risk-management, and technical-documentation processes.

Awards

Silver Medal

Project Title

TokenSaver: Context-Aware Semantic-Structured Document Retrieval System via Visual Memory

Winning Lingnan Faculty and Staff

·        Prof Li Zongxi, Assistant Professor of the Division of Artificial Intelligence of the School of Data Science

·        Xu Wangjiale, MSc student in Artificial Intelligence and Business Analytics

·        Li Yuxi, MSc student in Artificial Intelligence and Business Analytics

·        Li Yongxi, MSc student in Artificial Intelligence and Business Analytics

·        Wang Chen, MSc student in Artificial Intelligence and Business Analytics

Project Description

TokenSaver is an intelligent system specifically designed to assist AI in processing and answering questions based on long-form documents. It effectively addresses traditional AI shortcomings—such as high computational costs, vulnerability to irrelevant information, and the tendency to miss critical details buried in the middle—when reading vast amounts of text.

Key Features

·        Precision Content Targeting: The system addresses traditional AI limitations in comprehending complex charts and layouts, by combining a state-of-the-art visual language model with visual memory technology to interpret document structures and convert PDFs into machine-readable content. It then uses hierarchical semantic encoding and adaptive search functions to filter out irrelevant information and pinpoint key paragraphs, drastically reducing unnecessary data input.

·        Lightweight and High-Efficiency: The system cuts computational resource (token) consumption by 16 per cent on average while maintaining accuracy, and boosts search speeds by up to eightfold, providing a cost-effective, rapid, scalable knowledge management solution for handling massive volumes of professional documentation.

Awards

Silver Medal

Project Title

TransLab: AI‑powered Translation Learning and Evaluation Suite

Winning Lingnan Faculty and Staff

·        Prof George Chan Chi-yu, Assistant Professor of Practice of the Department of Translation

Project Description

TransLab is an innovative "AI + Education" platform designed for translation and language learning. By integrating AI-driven learning tools with pedagogical assessment tools, it significantly enhances both the acquisition and evaluation of translation skills.

The platform features two core modules tailored to support both student learning and teacher assessment: Transmuse, the learner module aimed at students and language learners, guides users through a two-stage learning journey from comprehension to refinement, first fostering a deep understanding of the source text, and then improving the quality of the translation. TransEval AI, the instructor module, is designed for educators and examiners, and evaluates translations across multiple criteria to highlight error types, pinpoint underlying causes, and offer context-specific suggestions for revision.

Key Features

·        Automated Assessment Workflow: The platform streamlines marking by automatically assessing student assignments, identifying error types and causes, and making suggestions for contextual revision. This increases teachers’ workflow efficiency, allowing them to dedicate more time to core teaching work.

·        Fostering Autonomous Learning: The system helps promote students' self-learning abilities, enabling them to steadily improve their translation and language proficiencies without direct teaching supervision.