Project supervisors: Mr Royston Shieh (shiehtw@nus.edu.sg), Mr Keith Tan (keithtcy@nus.edu.sg)

Industry partner/collaborator: CapitaLand Ascott Trust

Global hospitality operators depend on accurate, timely, and verifiable sustainability data from a large number of properties worldwide. These data underpin ESG reporting, corporate disclosures, and strategic sustainability decision-making. However, sustainability data collection today faces several challenges:

• The process is highly manual and time-consuming
• Data submissions are often incomplete, inconsistent, or late
• Regional differences introduce language barriers, varied local capabilities, and uneven understanding of sustainability requirements

While digital tools such as online forms, dashboards, and basic automation are already in use, sustainability teams at headquarters still spend significant effort on data chasing, validation, and reconciliation, limiting their capacity to focus on analysis and impact.

This capstone project invites students to design and prototype an AI-enabled toolkit or system that helps our industry partner to automate the extraction, validation, and transformation of sustainability data from known or input data sources. The system should intelligently map data into standardised formats suitable for dashboards, analytics, and sustainability reports, improving efficiency, accuracy, and scalability across global operations.

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

Industry partner/collaborator: HSC Pipeline Engineering Pte Ltd

Singapore's underground utility network is among the densest in the world. Before any excavation or road-opening works can commence, contractors are legally required under the Street Works Act and the Urban Redevelopment Authority Infrastructure Planning Authority Group (URA IPAG) guidelines to identify and map all underground utilities within the works boundary. For power cables specifically, this detection is performed using handheld electromagnetic locators (EML) — instruments that sense the 50Hz magnetic field passively emitted by live buried cables.

In current practice, a trained operator sweeps the EML instrument across the road surface and listens for a peak in the audible signal, then manually marks the cable position with spray paint. This process is inherently operator-dependent: detection quality varies with the skill and experience of the individual, and no data is recorded beyond a spray-paint mark on the road. If a cable is missed, there is no record of the survey coverage to indicate whether the area was scanned at all. Cable strikes during excavation — where construction equipment damages an undetected or incorrectly marked buried power cable — carry severe safety consequences including electrocution, supply disruption, and financial penalties. Despite advances in geophysical survey instrumentation globally, no continuous, data-logging, multi-sensor solution exists for this specific detection problem in Singapore's urban road environment.

The core challenge is that current instruments were designed to help an operator find a single cable at a point in time. They process the signal internally, apply automatic gain control, and output a locate point — not raw field data. As a result, the spatial magnetic field profile across the road cross-section is never captured, surveys cannot be audited or reproduced, and detection quality cannot be objectively measured.

This project explores whether a cart-based multi-sensor system capable of continuously measuring and logging the passive electromagnetic field across a road cross-section could provide a more systematic, auditable, and data-rich approach to underground power cable detection. Students are expected to investigate the physical principles governing the magnetic field emitted by buried live cables, design and prototype a sensor array capable of capturing this field with spatial resolution, and develop a signal processing approach that can identify cable positions from the logged data. The system should be testable in real road environments in Singapore, where HSC Pipeline Engineering Pte Ltd will provide access to field sites and ground-truth validation data from open-trench excavations.

Project supervisor: Mr Royston Shieh (shiehtw@nus.edu.sg)

Complex machines such as those used in lithography perform within very specific controlled conditions. Turbulent flow in cooling channels can result in performance degradation and disrupt the nanometer accuracy required in the industry. One of such contributing factors is the presence of bubbles in the cooling circuit. During the maintenance of such complex machines, the coolant must be drained and refilled in the shortest time possible to optimize the uptime of the machines.

The goal of this project is to introduce technical problem-solving skill by thinking beyond the object boundaries like a system engineer. In this holistic approach, the students will explore the basics of fluid dynamics followed by developing a process and design solution which will most effectively fill a cooling circuit with a standard detachable flange. The team shall develop a filling process, and make design improvements from the standard flange with given boundary conditions. The goal is to prevent and remove any trapped bubbles in the circuit. They are to build a simple test set-up with a monitoring system capable of identifying air bubbles, and validate their solution (filling process) and flange designs using the test set up.

Project scope:

• Explore the basics of fluid dynamics and their impact in cooling systems.
• Explore a variety of solutions and narrow down the options with help from design of experiment.
• Develop a process which minimizes bubbles creation during a filling process.
• Generate an FMEA for flange designs and test set up.
• Improve the design of a given standard flange to minimize/prevent trapping air bubbles.
• Build a simple test set-up capable of measuring bubbles and maybe vibrations.
• Test and validate their solution using the set-up to achieve a bubble free circuit within the shortest fill time.

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 computational audio, 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) or millimeter wave (mmWave) to localise the listener, these immersive experiences can potentially be automated and even follow a listener as they move around a room.

In this project (working with our industry partner Bang & Olufsen (https://www.bang-olufsen.com/en/sg), we will explore how to get the best positional accuracy using estimates from multiple UWB devices or combining with mmWave devices.

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

This 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 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.

Part of the goal of this project is to determine what AI technology is currently available to achieve the basic objective of the project and when technology may become available to continuously improve on it considering that the ultimate goal is to minimize human input as much as possible (i.e. manually creating question pools as opposed to having questions automatically created by the application).

Project supervisors: A/Prof Lim Li Hong Idris (lhi.lim@nus.edu.sg), Dr Kate Sangwon Lee (katelee@nus.edu.sg)

Industry partner/collaborator: Asian Women’s Welfare Association in Singapore (AWWA)

AWWA’s special education schools provide students with autism, aged 7 to 18, with nurturing and holistic educational experiences that build their independence and maximise their potential in tandem with their capabilities and aspirations. Children diagnosed with Autism Spectrum Disorder (ASD) often encounter significant challenges in developing the life skills necessary for independent living. Yet these competencies are critical for their long-term independence and quality of life.

Despite growing awareness of autism in Singapore and globally, there remains a shortage of accessible, scalable, and engaging tools to support life skills training for children with ASD. Key challenges include:

• Inconsistent training environments: Real-world environments such as MRT stations are unpredictable and can overwhelm children who depend on routine and structure.
• Lack of technology-driven solutions: While educational games exist, few are designed specifically for autistic children with a focus on real-world daily activities.

This project aims to design and develop a platform that is engaging, structured, repeatable, and accessible — one that mirrors real-life scenarios in a safe environment that children can navigate at their own pace.

The scope of the project includes the following:

• Study the needs of students with autism and select appropriate engagement tools.
• Design and develop a platform that simulates real-world daily life scenarios, with an initial focus on navigating the MRT system.
• Apply evidence-based principles of autism-friendly design, including predictable interactions, visual cues, low sensory overload, and positive reinforcement mechanics.
• Create a platform that is intuitive and usable by autistic children aged 9 to 16 without requiring constant adult supervision.
• Provide caregivers and therapists with a dashboard to monitor skill acquisition over time.
• Evaluate the user experience, stability and repeatability.

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

Predicting when flow in a horizontal pipeline transitions from stable stratified flow to potentially damaging slug flow is a critical design challenge in the oil and gas industry. Classical models such as Taitel-Dukler (1976) are widely used but were derived under simplifying assumptions that reduce accuracy at elevated operating pressures. There is a practical need for a fast, data-driven prediction tool that can generalize across a wide range of operating conditions and pipe geometries without requiring expensive experiments or simulations for every new case.

This project challenges the student to design, train, and validate a machine learning classification tool that predicts the stratified-to-slug flow transition boundary in horizontal gas-liquid pipe flow. The student will first conduct a structured literature review of flow regime transition physics and existing prediction methods, then compile a labelled training dataset from published experimental flow maps using dimensionless input features, liquid and gas Reynolds numbers, liquid Froude number, Eötvös number, and dimensionless operating pressure.

A baseline classifier (random forest or support vector machine) will be designed and tested in Semester 1, followed by refinement using a neural network architecture and systematic evaluation against held-out data and classical model predictions in Semester 2. The final deliverable is a validated, documented prediction tool with quantified accuracy and clearly identified operating limits.

Project supervisor: Mr Royston Shieh (shiehtw@nus.edu.sg)

Industry partner/collaborator: CapitaLand Ascott Trust

At the corporate headquarters (HQ) level, teams regularly collaborate with technology vendors to pilot new energy saving solutions and equipment across properties. These pilots are intended to validate whether the technologies perform as claimed and whether the resulting energy savings justify broader roll out across the portfolio.

However, reliably quantifying energy savings during pilot periods remains a major challenge. Energy consumption is influenced by fluctuating factors such as weather conditions and occupancy levels, and many properties lack comprehensive sensor infrastructure or historical baseline data for comparison. As a result, reported savings from vendors may not align clearly with observed changes in utility bills, making it difficult for HQ teams to make confident, evidence based scaling decisions. There is therefore a strong need for a practical, cost effective approach to energy savings quantification that works with limited data while remaining robust and transparent.

This project aims to design and prototype a cost effective, data driven methodology or decision support tool that enables HQ teams to quantify and validate energy savings from technology pilots under real world operating conditions. The solution should achieve the following:

• Enable fair and repeatable comparison of energy performance across pilot and non pilot periods
• Account for variability in influencing factors (eg climate, occupancy, and usage) without excessive manual effort
• Reduce dependence on expensive sensing infrastructure by identifying suitable proxies and alternative data sources
• Reconcile differences between vendor claimed savings and actual utility bill outcomes to support confident scale up decisions

Project supervisor: Dr Cai Shaoyu (shaoyucai@nus.edu.sg)

Smart wearables are transforming how we interact with the world, offering seamless and context-aware interfaces that integrate into everyday life. This project explores the design, development, and application of smart wearable technologies as ubiquitous interfaces that bridge physical and digital experiences. By embedding sensors, actuators, and communication modules into everyday accessories such as watches, glasses, clothing, or bands, these devices can perceive environmental cues, monitor user states, and deliver personalised feedback in real time. The project investigates key challenges in usability, comfort, data privacy, and multimodal interaction while prototyping wearable systems that enhance daily activities.

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

In industry the terms recruitment and talent acquisition are often used interchangeably referring to the requirement of filling an immediate or expected responsibility gap in the organization’s current employee profile. However, many human resource experts would suggest that while the two are related they are quite separate entities. Puja Lawanji defines recruitment as the process of filling a vacancy as the need arises while differentiating talent acquisition as a broader hiring strategy (Lawanii 2019). In fact, this concept of “a broader hiring strategy” summarizes findings in existing literature which typically suggests that the talent acquisition process is represented by multiple stages beginning with a talent requirement realization and ending with on-boarding the hired candidate. However, these many representations are often flawed by missing stages, convolution between primary and sub-primary steps and skewed representation to a particular industry sector.

One possible model suggests that the talent acquisition process includes the stages: Hiring Strategy, Value Requirements, Candidate Identification, Candidate Assessment, Candidate Selection, On Boarding and Retention and recognizes the subset of recruitment to include the stages of Value Requirements, Candidate Identification, Candidate Assessment and Candidate Selection. While there are many organizations that specialize in one aspect or another this overall process, very few consider it holistically. This project entails coming up with a product/service that considers the entire talent acquisition process to ensure that organizations are consistently recruiting and retaining the best candidates. It may include the development of one or more related applications that work in unison and consider each stage of the overall process.