Published Papers

This conceptual study critically examines the transformative impact of big data and artificial intelligence (AI) on organizational strategy, consumer engagement, and digital marketing. Drawing from a broad synthesis of recent scholarly and industry literature, this paper explores significant advantages, including enhanced personalization, predictive analytics, operational efficiency, and sustainability, alongside substantial challenges such as algorithmic bias, data privacy concerns, integration complexities, and workforce readiness. Emphasis is placed on the critical role of managers and stakeholders in mitigating technological risks, fostering ethical governance, and navigating complex regulatory frameworks. The study further highlights blockchain technology’s emerging role in improving data transparency, trust, and loyalty programs, thereby reshaping customer relationships. The paper provides comprehensive recommendations for managerial practice and policy, emphasizing adaptive governance, robust data stewardship, and collaborative innovation. By integrating these insights, organizations can leverage AI and big data responsibly to capitalize on their full potential, ensuring competitive advantage while addressing ethical and operational risks in an evolving digital landscape. 

This study examines industry-level valuation through the framework of the Capital Asset Pricing Model (CAPM) using monthly data for 49 U.S. industry portfolios from January 2000 to November 2025. First, a graphical analysis of total risk and return provides preliminary evidence of a positive risk–return relationship across industries. To isolate systematic risk, CAPM regressions are estimated to obtain industry betas and Jensen’s alphas. The Security Market Line (SML) is then used to assess whether industry returns are consistent with market risk exposure. The results indicate that while beta explains a substantial portion of return variation, several industries exhibit statistically meaningful positive or negative alphas, suggesting deviations from CAPM predictions. These findings highlight cross-industry differences in risk-adjusted performance and provide insights into industry valuation and the limitations of the single-factor model. 

Air pollution exposure varies significantly across micro-environments and directly impacts respiratory health, yet conventional monitoring approaches rely on sparse fixed stations that provide city-level measurements without capturing individualized exposure conditions or physiological responses. This lack of personalized monitoring limits the ability to assess how localized pollution affects individual respiratory health in real time. To address this gap, this paper presents LungHero, a portable multimodal sensing system that integrates localized air quality measurement with physiological respiratory indicators for personalized exposure assessment. The system combines low-cost particulate matter and gas sensors with temperature-humidity sensing, pulse oximetry monitoring, and acoustic cough analysis. A weighted Air Quality Index (AQI) is computed from normalized environmental sensor readings, while cough signals captured through a mobile device are analyzed using Mel-spectrogram features and a convolutional neural network to assess respiratory severity. Experimental evaluation demonstrates stable air quality sensing, effective cough classification with 89.44% accuracy, and observable correlations between degraded air quality, reduced SpO2 levels, and adverse cough patterns. The results suggest that multimodal sensing at the personal scale can provide meaningful insights into respiratory health under polluted conditions, offering a practical alternative to traditional station-based monitoring systems.

Efficient cellulose hydrolysis for bioethanol production is limited by its end product glucose as it inhibits β-glucosidases, the final enzyme in the multi-enzyme hydrolysis. Therefore, a glucose tolerant β-glucosidase with improved activity at high glucose concentrations is essential to overcome this product inhibition. Although a variety of such β-glucosidases have been identified or engineered, a kinetic model that can compare their performance under industrially relevant conditions is valuable for optimizing enzyme selection. In this study, a glucose-stimulated kinetic model was developed by extending an existing activation model to incorporate glucose inhibition effects. The model response was evaluated using previously reported β-glucosidase hydrolysis data of p-nitrophenyl-β-D-glucopyranoside (pNP-Glu). The resulting kinetic parameters, including glucose inhibition constants and maximum specific activity provide a quantitative framework for selecting suitable β-glucosidases for industrial bioethanol production.

The uses of Liquified Natural Gases (LNG) are vast, from generating electricity at huge scales for a town’s grid to being burnt using a gas stove to cook. LNG is just Natural Gas (NG) cooled down to -162°C (-260°F) to turn it into liquid form to make it easier to transport as, the volume has decreased (meaning more can be transported at once) and that it is in a liquid form which is more stable making it less prone to major disasters. Qatar is a major exporter of LNG, in fact it is the third largest producer of it after the US and Australia. Qatar produces 20% of the global supply, this means many countries rely on Qatar for their LNG. The main goal of this study is to build statistical models which can accurately analyse and predict Qatar’s LNG exports based on  previous data and benchmark variables. The data used in this study consist of Qatar's monthly export values for the period 2019-2024, the export volume and values in terms of the mass exported from Qatar for the same period, and finally the Henry Hub and Asia LNG prices as gas price benchmarks. The statistical and machine learning models used in this study are the ARIMA and the Random Forest models. The results show that while ARIMA provides a useful baseline prediction, the Random Forest model performs better because it can include more variables such as global gas prices. These findings suggest that machine learning methods can improve forecasting accuracy for LNG export markets. 

Smartphone ownership now commonly begins in early adolescence, raising concerns about dependency during a critical stage of social and emotional development. As smartphones can both support and hinder well-being, understanding adolescents’ own perceptions of dependency is essential. This study explores how American adolescents self-report smartphone dependency, perceived reasons for this dependency, and its emotional impact. Adolescents aged 14–18 living in the United States were recruited through targeted, random, and snowball sampling to complete a questionnaire on smartphone use and perceived dependency. The survey included open and closed-ended questions and Likert scales to assess emotional responses to phone-related scenarios and opinions about dependency in themselves and others. 18 adolescents participated, with 12(75%) identifying as female, ages ranging from 14-18. 94% of respondents believed others in their age group were attached to smartphones, only 44% reported feeling personally dependent. Participants reported an average smartphone use of five hours per day, with Instagram identified as the most frequently used app. Reasons for dependency varied, including social connection and the practical necessity of smartphones in an increasingly digitised world. Overall, respondents were hesitant to self-report dependency, despite recognising it as problematic in others, highlighting the complex nature of adolescents’ relationships with their smartphones.