Project supervisor: (to be confirmed)

Industry partner/collaborator: TE Connectivity

Problem statement:

• Manufacturing quality control systems struggle with accurate defect classification.
• Rare defects have limited training examples and are often missed.
• Manual retraining creates delays in detection and adaptation.

Proposed approach:

• Develop high-accuracy welding defect classification system.
• Create monitoring framework for defect pattern analysis.
• Implement automated model retraining pipeline triggered by pattern changes.
• Build visualisation tools for defect trends and system performance.

Potential innovations:

• Continuous learning system and adaptive machine learning architecture that responds to production changes.
• Anomaly detection for emerging defect patterns.

Expected/target benefits:

• Increased detection of rare defects leading to reduced flow out of defective products.
• Faster adaptation to new defect types and continuous improvement of detection accuracy.

Notes:

• Students who take up this project may have the opportunity to be engaged as interns during the project.
• Students who take up this project may be required to go for a site visit of TE Connectivity's manufacturing plant in Suzhou, China to better understand the problem and/or test their solution.
• Students who take up this project are expected to enter their project for the TE AI Cup in 2026.

Project supervisor: (to be confirmed)

Industry partner/collaborator: TE Connectivity

Problem statement:

• AI-based inspection systems require large datasets of defect images.
• Rare defects lack sufficient data for effective training and detection accuracy suffers for uncommon but critical defects.

Proposed approach:

• Develop generative models optimised for manufacturing defect generation with control mechanisms ensuring physical plausibility of different defect types.
• Implement systems to generate variations in orientations, lighting, and severity.
• Design validation methodologies for models trained on synthetic data.

Potential innovations:

• Generative AI with physical constraint mechanisms for realistic defect generation.
• Domain-specific data augmentation techniques.

Expected/target benefits:

• Improved detection of rare but critical defects.
• Balanced training datasets for AI inspection systems resulting in higher overall inspection accuracy.

Notes:

• Students who take up this project may have the opportunity to be engaged as interns during the project.
• Students who take up this project may be required to go for a site visit of TE Connectivity's manufacturing plant in Dongguan, China to better understand the problem and/or test their solution.
• Students who take up this project are expected to enter their project for the TE AI Cup in 2026.

Project supervisor: (to be confirmed)

Industry partner/collaborator: TE Connectivity

Problem statement:

• Manufacturing quality control currently lacks proper inspection capabilities.
• Labour-intensive manual microscope inspection for coil welding and relays is inefficient.
• Human inspection introduces variability and limits production throughput.

Proposed approach:

• Create AI image processing algorithms trained on production samples.
• Implement real-time processing system (>~1.5 seconds per piece).
• Establish a data collection framework that enhances AI model accuracy over time.

Potential innovations:

• Real-world application of computer vision and machine learning.
• High-resolution imaging in vibration-prone manufacturing environments.

Expected/target benefits:

• Increase production throughput.
• Reduce labour costs for manual inspection.

Notes:

• Students who take up this project may have the opportunity to be engaged as interns during the project.
• Students who take up this project may be required to go for a site visit of TE Connectivity's manufacturing plant in Trutnov, Czech Republic to better understand the problem and/or test their solution.
• Students who take up this project are expected to enter their project for the TE AI Cup in 2026.

Project supervisor: Dr Elliot Law (elliot.law@nus.edu.sg)

Industry partners/collaborators: Housing Development Board, South West Community Development Council

Recently, several town council estates in Singapore have stepped up efforts to control the population of pigeons by reducing human food sources and culling. Pigeons pose health risks as they carry diseases such as salmonella bacteria while their droppings can spread the ornithosis disease. These birds also cause damage to properties.

Pigeons are commonly found to nest on air conditioning ledges of residential buildings in Singapore. Current deterrents such as netting interferes with ledge functionality. Therefore, the goal of this project is to develop holistic solutions to deter pigeons from nesting on air conditioning ledges without hindering air-conditioner maintenance or clothes drying activities.

Students could explore:

• Technological solutions, such as targeted audio or visual deterrents that can react in a dynamic manner in the presence of pigeons.
• Physical design solutions, such as making the aircon ledges difficult for pigeons to land and nest.
• Maintenance-focused solutions, such cleaning mechanisms that allow residents to safely clear nesting materials from inside their homes.

The solution should also be:

• Weather resistant
• Cost-effective
• Easy to install
• Of minimal visual impact on building aesthetics
• Compliant with animal welfare guidelines

Project supervisor: A/Prof Mark De Lessio (mdeles@nus.edu.sg)

There is tremendous opportunity in the area of creating front end apps to interface with some of the Large Language Models that enable individuals and/or businesses to interact with them in a manner that adds value to both.  I would be very interested in exploring such opportunities with students that may have ideas that fit into this category.  The Automated Assessment Creation Tool project is based in the concept of how to fairly assess student comprehension based on the content delivery of the learning environment. The premise is the development of a device/app that can be used to record/scan delivered content and then interface with large LLM’s to automatically create assessments to measure the understanding of the content that has been delivered. There are a number of options to possibly achieve this through variations of AI ranging from doing key word assessment on the digitized content and then automatically determine the best assessment questions selected from a pool of pre-determined questions to actually developing the ability to automate the creation of relevant assessment material/questions based on the digitized content that has been captured. A significant goal is to ensure that the assessment is most closely based on the content that has actually been delivered in the actual classroom during the term.

Project supervisors: Dr Yen Shih Cheng (shihcheng@nus.edu.sg), Mr Graham Zhu (graham.zhu@nus.edu.sg)

Industry partner/collaborator: Bang & Olufsen

With the advent of spatial audio and 3D audio effects, smart speakers now have the ability to use advanced algorithms and audio processing techniques to place sounds in different parts of a room, creating immersive audio landscapes. In addition, by using precise positioning technologies like Ultra-Wideband (UWB) to localise the listener, these immersive experiences can potentially be much more directed and even follow a listener as they move around a room.

In this project, we will work with Bang & Olufsen (who are renowned for creating iconic audio experiences) to explore these and other technologies (e.g. on-device LLM for voice control (https://gektor650.medium.com/audio-transcription-with-openai-whisper-on-raspberry-pi-5-3054c5f75b95), parametric acoustic arrays (https://youtu.be/Fbphhg9ArXw), innovative mechanical designs to reduce discomfort while using headphones, actuating audio transparent materials in the Beosound Shape (https://www.bang-olufsen.com/en/sg/speakers/beosound-shape) to provide visual representations of music (https://youtu.be/VTEOu1p_0OU), etc.) to design and test new listening experiences.

Project supervisors: Dr Yen Shih Cheng (shihcheng@nus.edu.sg), Mr Graham Zhu (graham.zhu@nus.edu.sg), Ms Erin Ng (xin.le@nus.edu.sg)

Snoring and obstructive sleep apnea is a widespread sleep problem that affects millions of individual worldwide, disrupting their sleep quality and as well as that of their bed partner's sleep quality. To address this problem, this project aims to design an ultimate anti-snoring pillow that is comfortable, effective, and practical for integration into a user's daily life and customisable according to their preferences and sleeping posture. In addition, we will also be developing a simpler and cheaper way to measure and monitor sleep quality (apnea episodes) and positions of sleep.

Project supervisor: Dr Gabriel Lipkowitz (gel@nus.edu.sg)

Industry partners/collaborators: Apple

Spatial computing offers the potential to seamlessly blend the digital and physical worlds, opening up opportunities for qualitatively novel user experiences. In the context of physical product design, such technology offers the potential to bridge the divide between digital conceptualization and ideation, and those real-world contexts in which products will be applied by end-users. However, few applications have successfully demonstrated this concept in actual practice due to challenges in designing and developing applications and content for these platforms.

This project will seek to make progress towards realizing this goal by developing applications in the context of the burgeoning, yet still nascent, ecosystem around spatial computing, an industry priority. To do so, students will develop native to the Apple Vision Pro device using the high-level and general-purpose Swift programming language, implementing and drawing upon features such as 3D model interactivity, immersive scene understanding, and real-time simulation and rendering. Students will take specific contexts including but not limited to engineering, architectural design, and product design, and develop software applications to compellingly visualize such products in a user’s spatial environment, in a manner highly sensitive to user interaction and experience. Students will aim to release an app prototype on the Vision Pro App Store, which future work will extend to conduct in-depth human-computer interaction (HCI) studies.

Depending upon the project outcome, students will have the opportunity to pitch their projects to industry partners, potentially including at the Apple Developer Centre in Singapore, along with in the broader Singapore spatial computing development community.

Project supervisor: Mr Graham Zhu (graham.zhu@nus.edu.sg)

Industry partners/collaborators: Movement for the Intellectually Disabled of Singapore (MINDS), Down Syndrome Association (Singapore), Engineering Good Ltd

This project aims to support persons with intellectual disabilities (PwID) to perform the task of weighing baking ingredients, as part of vocational training and for inclusive employment jobs. Many PwID with Down Syndrome are literate and have difficulty processing numbers - they are not able to read recipes and understand for example whether a number like 500g is more or lesser than 450g. Most though are able to understand and follow verbal instructions.

To address this problem, this project aims develop a mobile app with the following features that can be paired with a weighing scale:

• Enable trainers to add in recipes
• User guided with step-by-step audio instructions
• Connect wirelessly with a weighing scale
• Determines whether required weight is reached and provides corresponding audio instructions

Students who take up this project will work closely with the industry partners and their clients or beneficiaries to co-develop and test the proposed solution.

Project supervisors: Mr Royston Shieh (shiehtw@nus.edu.sg), A/Prof Loh Wai Lam (mpelohwl@nus.edu.sg)

Multiphase flows are commonly encountered in the oil and gas, chemical, and power generation industries. The transition between different flow regimes — such as stratified, slug, annular, or bubbly flow — can lead to complex transient forces acting on the internal surfaces of pipelines and flow equipment. These flow regime-induced forces can result in vibration, fatigue, and failure if not properly understood and mitigated.

This project aims to investigate the hydrodynamic forces generated by different flow regimes in horizontal or near-horizontal pipe segments. Students will employ a combination of Computational Fluid Dynamics (CFD) and experimental methods at the NUS Multiphase Oil-Water-Air Flow Loop Facility to characterise and quantify these forces under varying flow conditions.

This project is suitable for Mechanical, Chemical, and Civil Engineering students who are interested in flow assurance. Students who wish to take up this project should have completed a course in fluid mechanics (e.g. ME2134, ME2135) prior to Year 4.

Project supervisor: A/Prof Mark De Lessio (mdeles@nus.edu.sg)

With rotary telephones tossed on the trash heap of a by gone era the concept of dialling has been replace by most users interfacing via a key pad with their mobile telephones. However, many people struggle with interfacing with their mobile phone in an efficient and quick manner because of the small size of the key pad used. This project proposes researching the availability of a technical solution that can be used to address this issue. For example, would it be possible to incorporate holographic technology or a similar technology to expand the key to enable an easier interface with the key being selected as it is approached by the finger/thumb doing the selection. Perhaps there is an alternative technology that could be incorporated to achieve the same result.

Project supervisors: Dr Aleksandar Kostadinov (aleks_id@nus.edu.sg)

Background

Timely, localized plant health data is critical for reducing input waste and improving crop yields. Biosensors can provide real-time, in-situ monitoring of plant stress, nutrient levels, or disease, yet their integration into field systems remains underdeveloped.

Aims

To explore, design, and test low-cost biosensor concepts for detecting key plant health indicators (e.g., soil, water, and ambient properties; plant health; pest and pathogen presence).

Methodology

This research-oriented project will involve a survey of current biosensor technologies, selection of target biomarkers, and conceptual design of sensor units. Depending on scope, lab testing or simulation of sensing mechanisms may be conducted. Collaboration with researchers from various departments within NUS will be encouraged in which existing technologies can be used for further development.

Deliverables

• Technology review
• Biosensor design concepts
• Proof-of-concept setup (lab or simulation-based)
• Feasibility analysis
• Research paper-style report and showcase poster

Note:

The proposed topic, methods, and deliverables are tentative and intended as a starting point. Depending on the team’s size, interests, and strengths, both the direction of the project (research or development) and its specific outputs (hardware, software, principles) can be adapted accordingly. Students are encouraged to explore related avenues if they find them more compelling.

To support this process, a selection of handbooks and manuals is provided to help kickstart research and frame initial explorations:

• Biosensors in Agriculture: Recent Trends and Future Perspectives | SpringerLink
• Handbook of Precision Agriculture | Principles and Applications | Anch
• Nanosensors for Agriculture, Water, Environment, and Health | SpringerLink

Project supervisor: Mr Nicholas Chew (nickchew@nus.edu.sg)

Industry partner/collaborator: Epson Robotics

In our tech-driven world, automation and robotics are revolutionising various industries, including food and beverage services. The traditional French press coffee-making process, which takes about 10 minutes to prepare just 4 cups, is an ideal candidate for automation. This project aims to optimise the coffee-making process by improving the motion path planning of two Epson 6-axis robots, thereby reducing cycle time and enhancing operational efficiency. A demo machine was previously developed for an event at the Singapore Museum. This project offers an exciting opportunity to refine the coffee-making process, further develop this robot-based system, and create a solution that can be showcased both locally and internationally at exhibitions, promoting the future of "manless" coffee making.

Project Objectives:

1. Reduce Coffee Preparation Time: Streamline the process and reduce the preparation time to just 1-2 minutes.
2. Automation and Stable Operations: The machine must operate autonomously and ensure consumables (paper cups, water, coffee beans, etc.) can be stored for up to 3 days of continuous coffee-making without requiring human intervention.
3. Network Architecture Improvement: Removes the BnR PLC and transitioning control to the robots themselves. Digital inputs and outputs will be connected to each robot controller to establish the necessary communication handshake between the machines.
4. Explore Innovative End-Effector Designs: Students are encouraged to think creatively and come up with new and innovative end-effector designs to streamline the coffee-making process, improving overall system efficiency.
5. Robot Specifications: The system will be limited to two Epson 6-axis robots, each with a 4 kg payload capacity and a 600mm arm reach, making it necessary to optimize the design and operational processes within these constraints.

Project supervisor: A/Prof Mark De Lessio (mdeles@nus.edu.sg)

There is tremendous opportunity in the area of creating front end apps to interface with some of the Large Language Models that enable individuals and/or businesses to interact with them in a manner that adds value to both.  I would be very interested in exploring such opportunities with students that may have ideas that fit into this category.  The “Smart Advisor Tool” consists of  a front end tool that assist individuals to assess what their personal interest are and how that might correlate to the best degree programs that would enable them to realize these interests as working adults.  This would essentially leverage AI to categorize their interests, identify relevant degree programs, identify the types of jobs available to individuals with similar interests, identify the best universities that offer such programs and eventually even perhaps the organizations that have such jobs available.  There would be many ways to monetize such a tool including working with Universities to advertise their programs and/or partnering with job placement agencies or even with organizations themselves to push potential candidates to them that have already been vetted and matched with positions they currently are seeking to fill.