This project develops a novel HPV detection platform to advance cervical cancer elimination per WHO 90-70-90 targets, boosting screening access, enabling early accurate detection and timely treatment to cut incidence and mortality—aiming for 70% screening coverage for women aged 35/45 and scalable solutions for underserved areas. Theoretically, it informs advanced diagnostic tech application in global health and HPV/cervical cancer epidemiology in low- and middle-income countries; practically, it validates integrating cutting-edge tech (point-of-care testing, cloud data) into healthcare systems, with lessons for other disease diagnostics. Its innovation lies in a portable, affordable AI-integrated next-gen HPV platform supporting same-day diagnosis, real-time data connectivity for centralized monitoring/targeted vaccination, and equitable access via mobile clinics/community workers. DKU’s public-private collaboration with Shanghai Annealing Biotechnology is key to its feasibility.

Global population ageing and the high prevalence of chronic diseases call for family-centered health management, yet current research and interventions have critical limitations: an adult-centric focus that overlooks children/grandchildren’s meaningful role in family health, one-dimensional framing of intergenerational health communication as one-way support, age-siloed health education design, and underutilization of narrative, co-creation-based health communication methods and tailored digital tools—creating key gaps in both theory and practice.
To address these gaps, this project redefines chronic disease management as an intergenerational, relational and communicative process. It adopts a participatory co-creation approach integrating multiple disciplines, engaging older adults and children to develop age-appropriate, culturally responsive health communication tools (e.g., picture books, storytelling kits, low-threshold digital products) for common chronic conditions and core lifestyle domains. The initiative advances relevant theories, delivers scalable practical tools for diverse settings, and enables student-faculty collaboration to drive both scholarly and translational impact.

This project leverages Chinese tea culture as an experiential platform to foster intercultural learning, identity reflection and student-driven innovation at Duke Kunshan University. Building on existing tea culture initiatives and the student-led Tea Culture Club, interdisciplinary student teams design, deliver and evaluate tea-integrated cultural innovation activities, linking traditional tea culture with contemporary campus life, identity building and intercultural communication.
Adopting a learning-by-doing approach, students act as designers, organizers and researchers – planning tea workshops, roundtables, pop-ups and markets, while conducting research to track participants’ cultural learning outcomes. This builds their intercultural collaboration, project management, cultural communication and applied research skills.
In later stages, students translate experiential learning into cultural product prototyping and creative incubation (e.g., tea branding, bilingual storytelling, student-designed tea kits), exploring the path from embodied cultural experience to creative development and social innovation.
Integrating academic inquiry and campus practice, the project explores how embodied cultural practices support intercultural learning, builds a sustainable student-led tea culture innovation model for DKU’s campus, and its outcomes may be further developed into Independent Study or Signature Work.

Taxicab services are vital to urban mobility, and optimizing route recommendations is key to boosting driver income, passenger experience and traffic efficiency in major cities. Unlike traditional single-driver optimization, modern ride-hailing demands multi-user route recommendation in dynamic shared traffic environments, which faces challenges of high computational complexity, heterogeneous objective coordination and the need to balance overall efficiency with driver fairness.
Multi-agent reinforcement learning (MARL) is a natural solution but current methods suffer from poor scalability, unfair recommendations or inefficient communication mechanisms. This project develops a scalable, adaptive MARL-based taxi route recommendation system to balance efficiency, fairness and computational cost, with four research stages: formalizing the problem with fairness-aware metrics; designing adaptive agent communication/coordination mechanisms; validating the approach via large-scale real taxi trajectory data against state-of-the-art baselines; and building a practical application for real-time, long-term route recommendations across different temporal conditions.

This project develops a China-tailored multi-LLM orchestration framework for Gaokao preparation, enabling dynamic coordination of multiple language models to solve complex exam problems and deliver personalized university/major recommendation support. We partner with a leading Turkish AI education startup—whose university entrance exam platform serves 500,000 active users with mature AI problem-solving and major preference analysis functions.
Leveraging the startup’s validated technical architecture and under its technical supervision, we first build a localized prototype: collecting and processing Chinese educational data to adapt the system to Gaokao question formats, Chinese linguistic features and local educational needs, while validating data methodologies, user interaction and baseline performance in the Chinese context.
Upon prototype validation, we integrate our core innovation: a China-specific multi-LLM orchestration framework. Distinct from single-LLM approaches, it intelligently allocates tasks to Chinese educational data-optimized heterogeneous LLMs based on problem complexity, subject domain, computational constraints and pedagogical requirements.
This architecture drives key enhancements: cost efficiency (lightweight models for simple tasks, advanced models for complex reasoning), higher accuracy (matching tasks to domain-specialized LLMs), lower latency (rational workload distribution), and balanced pedagogical effectiveness and computational resource utilization.
The research upgrades the prototype into a technically differentiated system, addressing practical challenges in China’s educational AI landscape with our LLM coordination innovations and building a robust technical foundation for future EdTech applications.

This project develops an automated algorithm design (AAD) framework for unmanned aerial vehicles (UAVs), generating mission-tailored designs by automatically synthesizing and adapting optimization strategies. It optimizes airframe geometry, sizing and key subsystem parameters for specific tasks such as long-endurance surveillance, cargo delivery, and urban low-noise flight, centered on software-level development integrating aerodynamic, performance and mission profile computational models with AAD methods. Initially focusing on virtual conceptual design, optimization and benchmarking in simulation, it ensures geometry compatibility with downstream manufacturing workflows for potential 3D printing or conventional fabrication.
The project begins with a prototype phase using established UAV design test cases, defining standardized parametric configuration spaces and building an automated evaluation pipeline for mission performance metrics—laying the foundation for AAD experiments that replace manual algorithms with auto-synthesized strategies. Its core innovation is the AAD framework itself: instead of fixed algorithms, it automatically designs and adapts strategies by composing and tuning core components based on benchmark performance feedback, orchestrating a portfolio of auto-designed algorithms for rigorous comparative analysis.
This framework advances UAV design and optimization research by shortening design cycles, improving quality via algorithm-tailored multi-objective optimization (performance, robustness, manufacturability), and providing a systematic basis for comparing algorithm families. In the long term, its 3D geometric representations enable transferring optimal virtual designs to CAD tools for physical prototyping and validation when resources permit.

This project develops a student-built Business Chinese mini-program and mobile app to address the gap in practical workplace Chinese learning for global professionals, as existing platforms focus on general language and test content rather than authentic workplace communication (e.g., meetings, negotiations).
Interdisciplinary student teams, partnering with companies, will conduct workplace needs analysis to create a market-oriented product. Under faculty supervision, they will design learning pathways and refine the platform, with core content covering four high-frequency professional communication modules via scenario-based lessons and pragmatic guidance.
Following a learning-by-building approach, students lead all aspects from user research to MVP development, gaining interdisciplinary and EdTech experience. Outcomes include a functional prototype, instructional modules, and testing reports, with potential for further development into academic or entrepreneurial initiatives.

Music is a powerful tool for emotion and social connection, yet Deaf and Hard-of-Hearing (DHH) people face barriers to experiencing it due to its auditory nature. Current AI music visualization systems link music information retrieval with visual synthesis, but fail to convey music’s expressive, dynamic qualities in an embodied, socially engaging way.
We propose a low-latency music visualization system driven by a robotic arm, which converts audio features into UV light trajectories projected on a phosphorescent surface. Unlike virtual visualizations, the robotic arm turns auditory signals into tangible kinetic motion—perceivable visually and physically—while light visualization adds multimodal feedback, mapping rhythm, loudness, timbre and other musical features to dynamic visual cues.
Rooted in the idea that music can be seen and felt as well as heard, the system creates inclusive, immersive music experiences for DHH individuals. It enables them to connect with music personally and participate in social musical activities like dancing, fostering a stronger sense of belonging and breaking the historical exclusion of DHH people from musical and social worlds.

This project explores generative music as an alternative form of journaling, where personal reflections, moods, or daily experiences are translated into evolving musical compositions. Instead of (or in addition to) writing text, users “journal” by shaping musical parameters—such, such as tempo, harmony, texture, or instrumentation—through prompts, interactions, or data traces. The goal is to investigate how generative music can support reflection, emotional expression, and self-awareness, and how it compares to traditional text-based journaling.Explore non-textual journaling through sound and musicInvestigate how generative systems can encode emotion, memory, or experienceExamine the role of interaction and control in reflective practicesReflect on user experience, expressiveness, and meaning-making

This project aims to produce studio-quality recordings of Kunqu singing and to develop a research-informed approach to reaching listeners outside China. Undergraduate students will be involved throughout the process in three connected strands: (1) recording and post-production of selected Kunqu arias with senior singers and minimal accompaniment; (2) international audience and platform mapping that examines how Chinese classical music and related genres are packaged and discovered abroad, including genre labeling (for example world music, Chinese music or classical), production style, instrumentation choices, release formats and platform ecosystems; and (3) translating that research into a distribution and outreach plan that is piloted in practice, including bilingual contextual materials (track notes, listening guides, short explainers), platform-ready assets and targeted outreach to the communities most likely to listen.The innovation lies in treating heritage recording not only as preservation, but as a global-facing communication problem: how to record, package and contextualize a stage-based form so it can live as audio in international contexts (streaming platforms, digital archives, educational use, curated channels), rather than being channelled primarily through China-specific cultural heritage or tourism framings.

This project explores how a small pure-acoustic indoor theatre based on traditional Chinese architectural methods could be designed and made viable in partnership with a local cultural development zone. China currently lacks venues that are purpose-built for unamplified traditional performance and designed according to both acoustic science and traditional architectural principles. Students will investigate how that gap might be addressed in practice, working with a local government-directed project developing a theatre village in the area near DKU.Students will: (1) study a small number of benchmark venues across the world that support unamplified performance and summarize the basic spatial and acoustic ideas that make them work; (2) map potential users, partners and revenue streams for a small acoustic theatre in Kunshan (3) help prepare a concise slide deck and take part in a live pitch to both investors and the development zone partner during the project period. The project connects research on performance space with practical questions of business and partnership models, and gives students experience in taking a cultural concept from initial research to an actual proposal to potential backers.