Ranked 1st place in the 3rd International Data Science Olympiad (IDSOL 2024)
Also: 1st place at Emerging Technology Olympiad (ETO) at DigiQuest 2024

Adolescent Idiopathic Scoliosis is the most common spinal disorder, affecting up 130 million teenagers. With unknown root causes, early detection and intervention are critical, but traditional methods of scoliosis screening require medical professionals, specialized equipment or radiographic exposures, making mass screening inconvenient and inefficient.

This study aims to address this problem by developing a fast, convenient screening method that can detect potential scoliosis from bare back images. A product prototype was built, utilizing K-Nearest Neighbor (KNN) with 72% accuracy and a Convolutional Neural Network (CNN) with 77% accuracy.

This initial validation highlights the product’s potential for mass scoliosis screening and ongoing patient progress monitoring during scoliosis treatment. Business model, governance model, scalability, further model improvements and pending clinic data collection to increase dataset are discussed.

Ranked 2nd place in the 3rd International Data Science Olympiad (IDSOL 2024)
Also: 5th place at Emerging Technology Olympiad (ETO) at DigiQuest 2024

A vast majority of people in developing regions lack medical care due to poor communication infrastructure, high cost of medicine, or a dearth of medical professionals. Moreover, a lot of organizations lack support for those facing non-emergent but life-altering chronic illnesses. There are organizations such as Doctors without Borders, WHO, and Partners in Health trying to help with remote medicine. However, none of them can provide long-term medical advice, nor are they able to service the entire population consistently.

To make medical services scalable to those in need, we developed a device with computer vision capabilities that will intake a picture taken from the device and output traditional Chinese medical advice. In particular, we focus on how tongue diagnosis can be developed. It can diagnose and offer medical advice by applying computer vision to observable symptoms. By building upon theories of traditional Chinese medicine (TCM), we make the user experience much more user-friendly and intuitive. In particular, our prototype focuses on the technical processes that are required in the autonomous tongue diagnosis of TCM.

Ranked 3rd place in the 3rd International Data Science Olympiad (IDSOL 2024)

Though life planning education has been launched for over 10 years, many secondary schools currently lack a comprehensive and holistic approach. Students often view life planning as outdated career programs, with disorganized new activities like job shadowing added without clear purposes. The dominant academic culture in many especially Asian countries also renders other learning unimportant, resulting in a lack of student motivation and an incomplete understanding of their strengths, needs, interests, and potential.

The "Potential+" app aims to address these issues by providing a more integrated platform for life planning education. It facilitates the collection and analysis of comprehensive student records, including academic, non-academic, quantitative, and qualitative data, along with personality identification and target studies/ careers. The app also helps evaluate students’ generic skills and matches them to suitable learning activities and target studies/careers. Applying generative AI, the app also provides personalized recommendations to guide students in preparing for their future studies and careers.

Furthermore, "Potential+" uses generative AI to assist teachers and students in efficiently and accurately preparing personalized documents, such as reference letters and personal statements. By taking a more holistic approach to life planning, the app seeks to cultivate a growth mindset and encourage diverse learning pathways, ultimately leading to a more fulfilling school experience for students and producing a more productive and fulfilling workforce for society.

Award of Merit in the 3rd International Data Science Olympiad (IDSOL 2024)

In many parts of the world, students from different communities face unequal access to education due to economic, social, and cultural factors. However, modern technology has the potential to bridge this gap and promote greater educational equity.

Many ed-tech platforms have tried to bridge the gap, but have failed due to static content and audience bias. For the content to be actually effective for students, we need to be able to cater to learners with different learning capabilities and learning types. At the same time, content generation should also be focused on for effective education transfer.

StudyGazelle is a combination of systems that streamline the process of learning content generation and distribution with the power of Large Language Models.

We propose a breakthrough in adaptive learning through streamlining the process of creating Pedagogical Conversational Agents (PCA). We also showcase the capabilities in pedagogical approaches based on different learner types through our user platform ”StudyGazelle”.