A matrix equation asks a simple question: given F(X) = 0, what matrix X solves it? This talk looks at this question through several examples, including the square root of a matrix and quadratic matrix equations that appear in vibrating systems such as bridges and buildings. It also looks at matrix equations that describe population growth, and newer work on matrix and tensor functions, an active and growing area of research. These ideas connect closely to artificial intelligence, since a neural network is built from matrices, and training it means solving a large optimization problem. This links matrix theory, algorithms, and optimization directly to the themes of this conference. The talk then moves from theory to real use: pricing financial products through simulation, planning sustainable fishing over several years, modeling eye shape for automatic medical image analysis, and reducing waste in factory production. Together, these examples show that industrial mathematics is not a small side branch of theory, but a natural next step, where mathematics meets real industry problems and mathematicians must be ready to try new tools without fixed assumptions.

Breast cancer remains one of the most challenging diseases to model due to the complex interactions among tumour cells, immune responses, metabolic regulation and therapeutic interventions. Understanding these interconnected biological processes requires mathematical frameworks that not only describe disease progression but also support prediction, optimisation and intelligent decision-making. This keynote presents a controloriented mathematical modelling framework that integrates tumour growth, natural killer (NK) cells, CD8+ T cells, glucose-mediated metabolic dynamics, immune suppression, treatment delay and parameter uncertainty within a unified delay differential equation model.

The presentation demonstrates how advanced mathematical modelling can reveal the dynamic mechanisms governing tumour progression, immune surveillance and immune escape while accounting for biological uncertainty through fuzzy modelling. The incorporation of optimal control theory provides a systematic approach for identifying treatment strategies that minimise tumour burden while improving therapeutic efficiency.

Numerical investigations further illustrate how hyperglycaemia, delayed immune activation and treatment timing influence long-term disease outcomes and therapeutic effectiveness.

Beyond the mathematical analysis, this keynote highlights the emerging role of artificial intelligence in mathematical oncology, where AI-driven parameter estimation, model calibration, predictive analytics and personalised treatment optimisation complement mechanistic mathematical models. The integration of mathematical modelling, optimal control and artificial intelligence offers a promising pathway towards precision medicine and intelligent healthcare systems. More broadly, this work illustrates how industrial mathematics can bridge mathematics, biomedical science and AI to address complex realworld healthcare challenges and support evidence-based clinical decision-making.

Malaysia’s ambition to become a Negara AI is not one institution’s task, but a whole-of-nation effort spanning government, industry, and academia. This keynote traces that journey through the National AI Action Plan 2026 to 2030 (Negara AI 2030) and the evolution of the National AI Office into AI Malaysia, connecting national policy to the mathematical and statistical foundations industry needs to turn AI ambition into deployable, trustworthy systems. Drawing on Malaysia’s experience building AI capability under real institutional constraints, it offers a candid view of what whole-of-nation AI coordination actually demands.

This paper presents a novel framework integrating machine learning (ML) and mathematical optimization for managing agricultural waste in a community enterprise focusing on coconut production. A hybrid approach is proposed to tackle the seasonal nature of biomass supply. First, an ML model is developed to predict the generation of coconut by-products (husks and shells) from a 40-acre plantation. Second, the predicted values serve as parameters for a Mixed-Integer Linear Programming (MILP) model designed to optimize the allocation of biomass across various processing alternatives, maximizing economic value while adhering to operational constraints. The preliminary result shows that the proposed model provides a superior decision support tool for achieving zero-waste goals in local green industries, showcasing the practical application of AI-powered industrial mathematics.

The Multivariate Exponentially Weighted Moving Average (MEWMA) control chart is widely used for monitoring multivariate processes because of its ability to detect small and moderate shifts more effectively than conventional control charts. However, the performance of the classical MEWMA chart depends on the accurate estimation of the process mean vector and covariance matrix, which can be severely affected by non-normality and outlying observations commonly found in industrial and environmental data. This study proposes a robust MEWMA control chart based on the Fast Minimum Covariance Determinant (Fast MCD) estimator. By providing robust estimates of multivariate location and dispersion, the Fast MCD estimator reduces the influence of extreme observations and improves monitoring performance under non-normal process conditions. A smoothing parameter of λ = 0.3 was employed to enhance the sensitivity to gradual process shifts while maintaining robustness against data contamination. The novelty of this research lies in integrating the Fast MCD estimator within the MEWMA framework for the simultaneous monitoring of three critical water production quality characteristics: turbidity, pH, and residual chlorine. The proposed methodology was applied to operational water quality data consisting of 331 and 121 observations in Phases I and II, respectively. After removing out-of-control observations, Phase I successfully established a stable in-control baseline with an upper control limit (UCL) of 13.79. Using this robust reference model, Phase II monitoring identified 47 out-of-control signals, indicating potential process deterioration and quality variations. The results demonstrate that the Fast MCD-based MEWMA chart provides reliable control limits and effective detection of multivariate process shifts, making it a valuable early warning tool for water quality assurance and operational decision-making.

The increasing complexity and digitalisation of New Energy Vehicle (NEV) supply chains require effective approaches to identify and prioritise emerging risks. This study develops an integrated approach combining Transformer-based text classification with Analytic Hierarchy Process (AHP) and fuzzy comprehensive evaluation for NEV supply chain risk analysis. First, 1,000 English open-source risk records were classified into six categories: operational, supply, logistics, information and data security, financial and inventory, and environmental/ESG/compliance risks. The dataset was developed through a two-stage human annotation and adjudication process. BERT and RoBERTa were evaluated using standard fine-tuning, class-weighted loss, and focal loss. Results demonstrate that Transformer-based models can effectively organise unstructured risk information into structured categories, with RoBERTa using standard fine-tuning achieving the strongest performance. Class-weighted loss improved several configurations, while focal loss showed no consistent advantage. Second, AHP was applied to prioritise the identified risk categories and underlying factors based on expert judgement, followed by fuzzy comprehensive evaluation. Information and data security risks received the highest priority, followed by supply and operational risks. At the secondary level, BMS-related failures, supply shortages or disruptions, software instability, critical material dependence, and component-level defects were ranked among the most important risk factors. These findings highlight the importance of digital reliability, supply continuity, and battery-related technical stability in NEV supply chains. By combining automated risk signal screening with expert-based prioritisation, the study provides a structured methodological basis for converting unstructured risk information into actionable risk intelligence and supporting more systematic NEV supply chain risk monitoring.

In this talk, I will present examples from materials science and chemistry showing how AI methods combined with human scientific expertise can support the design, optimization, and control of advanced materials and chemical systems. The focus is on approaches that remain effective in small-data regimes by incorporating physical intuition, structural modeling, and domain knowledge.

Rather than relying on large datasets, these methods integrate modeling and scientific understanding to improve interpretability and robustness. Case studies will include molecular machines, and related systems, highlighting how mathematics, AI, and domain knowledge together enable new opportunities for discovery and control.

Quantum machine learning (QML) is increasingly explored as a framework for modelling complex, high-dimensional systems characterised by nonlinear interactions and rare events. In parallel, cyber incidents are being recognised as potential sources of financial instability as digitalisation intensifies interdependence between technological infrastructures and financial markets. Existing monitoring tools typically analyse cyber and market indicators separately, limiting their ability to detect early stage instability arising from their interaction. This paper proposes a conceptual hybrid quantum–classical anomaly detection architecture for cross-domain systems, with cyber–financial stability serving as a motivating application. The framework integrates cyber and market indicators into a unified representation through quantum feature mapping, with the potential to model nonlinear dependencies that are difficult to capture using classical methods. A Grover inspired amplitude amplification mechanism is incorporated to enhance the identification of rare or subtle instability signals indicative of emerging cyber–financial stress.

Drawing on established financial stability frameworks from the IMF and CPMI–IOSCO, the proposed architecture illustrates how quantum-enhanced learning can support early-warning analysis in interconnected cyber–physical systems. As a conceptual study, this work establishes a foundation for future empirical validation, simulation-based stress testing, and the development of quantum enhanced monitoring and supervisory tools aimed at strengthening financial stability in an increasingly digital and interconnected environment.