This paper presents a computational design framework for evaluating and optimizing passive thermal comfort strategies in tropical residential buildings, with a focus on Kuala Lumpur, Malaysia. Through a synthesis of climate data, building performance literature, and simulation-based analysis, a structured set of design principles is developed across three scales: macro (urban context), meso (building form), and micro (material and detail). Nineteen performance metrics drawn from Malaysian building standards and academic research guide design decisions at each scale, covering indicators such as the Urban Heat Island effect, Predicted Mean Vote, wind speed, and Energy Use Intensity. The framework adopts a static thermal comfort model and a performative design approach, enabling architects to test passive strategies including natural ventilation, solar shading, and material selection before construction. Results indicate that early-stage computational simulation can significantly reduce reliance on mechanical cooling while improving indoor and outdoor thermal comfort. Although developed for a Malaysian context, the framework offers transferable principles for other hot and humid climates, and lays groundwork for future integration of adaptive comfort modeling and AI-driven optimization.

With an increase of data documentation and standardization in the construction field in The Netherlands, by norms such as the NEN, there is a possibility to introduce data-driven approaches to certain areas within the construction industry. One of these is the area of budget estimation which is currently fully dependent on a cost estimating professional. Due to the need for estimations that are effective and time-efficient, especially in the primary phase of a project, the potential of introducing a data-driven approach is explored through this thesis. The main objective of this research is the development of a data-driven model, in the form of a Virtual Assistant (VA), to increase the objectivity of the estimation of maintenance budgets of civil engineering structures. From a literature study it is apparent that a fitting data-driven approach for the development of this model is the machine learning technique Decision Tree Classification (DTC). The VA model is developed using historical data, in the form of past input and past output, to train the model and therefore make predictions. Data that is used as past output for this model is a budget range which is documented as a budget class and data that is used as past input is data that is ensured to be objective and gives a description of each bridge. In this case the past input data are the characteristics of the bridge which refer mostly to the dimension of the bridge, the NEN2767, which captures the decomposition and condition of the bridge and to a lesser extent the duration of the maintenance. Through exploring past cases the machine learns rules and predicts the outcome for a new case and therefore predicts the budget range. This shows in which range the budget guess of the estimator should fall. Generally in order to develop a VA model and make it applicable for industry use it is important that an organization that uses such model aligns their data storage with the DTC methodology. This means the introduction of standardizing data in classes and the introduction of standard procedures to document the data. Only when these elements are present within the organization, the data that is used for past input and output can be regarded as objective and the VA can fulfill its function which is to verify the budget estimators guess in an objective manner.