Project proposer: Hu Lifan (e1480030@u.nus.edu)

This project seeks to empower visually impaired individuals with an intuitive, AI-driven understanding of their surroundings. By integrating cutting-edge Vision-Language Models (VLMs) and embedded AI hardware, the system will describe environments in real time — helping users identify people, locate important objects, and navigate their daily spaces safely and independently.

The goal is to design a lightweight, wearable assistant that operates fully on-device, ensuring both privacy and low latency. Beyond mere object detection, the system aims to convey meaningful context: recognizing familiar faces, describing room layouts, and detecting potential hazards. The device will be designed for ease of use, allowing users to operate, charge, and maintain it without external help.

Through this project, we will develop a hardware-software pipeline, encompassing perception, interaction, and accessibility design. Ultimately, this project aspires to restore autonomy, dignity, and confidence to visually impaired users—bridging human perception and intelligent computing through accessible technology.

Project proposer: Tsai Ho-Hsun (e1482086@u.nus.edu)

Motivation: Introduce/develop more autonomous systems in a B2C setting.

Intended objective: Automated system implemented in beverage shops to reduce labour cost and increase available hours.

Deliverables: A functioning system with a kiosk that can be implemented in a shop directly with routine restocking of raw ingredients, and a report analyzing the pros and cons of this automated system with existing method.

Project proposer: Tay Hock Jun (e0702171@u.nus.edu)

Lymphedema is a non-curable chronic condition affecting 250 million patients worldwide that results in accumulation of lymphatic fluid in the limbs. Our proposed solution, FREESTRYDE, is a lightweight, modular, active compression device designed for everyday wear. It contains pressure sensors to monitor swelling and pulls heart rate and bioimpedance data from a smartwatch or fitness tracker to monitor user activity. It then uses AI to predict swelling patterns based on user activity and generate specific compression profiles which apply adaptive compression before discomfort starts and only when necessary. The compression mechanism is a mechanical band tensioning system which provides peristaltic compression to the limb to drive accumulated fluid upwards. Both clinicians and patients can track progress real time via our connected app, reducing unnecessary clinic visits.

Project proposer: Magdalene Lim Yong Qi (magdalenelim@u.nus.edu)

This project aims to explore how smart wearable technology can be seamlessly integrated into the daily lives of the elderly (eg. wearable smart socks). The primary objective is to prevent falls and enable rapid response and recovery among community-dwelling seniors. This initiative will be conducted in collaboration with our stakeholders at the Tan Tock Seng Hospital (TTSH) Innovation Department.

Through this project, we will develop an integrated hardware–software system that encompasses perception, interaction, and accessibility design. Ultimately, the project seeks to bridge the gap between patients and the healthcare system, enhancing visibility, responsiveness, and quality of care.

Project objective: design a familiar, low-adaptation wearable that seamlessly integrates everyday comfort with smart functionality. Equipped with real-time fall detection and alert systems, the device will provide continuous, reliable monitoring to enhance safety and peace of mind for both elderly users and their caregivers. Prioritising ease of use and accessibility, the design will be suitable for both home and hospital environments, supporting not only older adults but also secondary users such as community nurses, caregivers, and allied health professionals.

Project proposer: Neo Yu Xuan (e1157210@u.nus.edu)

Waste Too Taste aims at using AI-driven recommendation logic — similar to how Netflix suggests shows — to smartly pair surplus/near-expiry supermarket items into meal kits that make sense together. By analysing item attributes, purchase patterns, and recipe databases, the AI identifies complementary ingredients and bundles them into optimised “blind bags,” helping supermarkets clear stock efficiently while offering customers convenient, ready-to-cook sets. The end product would likely be a software that can be used easily by supermarket staff. The project can even be potentially expanded towards other target audiences (e.g. hawker stores struggling with surplus ingredients at the end of the day could utilise this food-pairing tech to come up with new recipes that they can sell as a "surplus meal").