In modern chemistry, lipophilicity, which refers to the affinity of a compound for lipids or fats, is recognized as one of the crucial parameters influencing the behavior and properties of organic compounds. This parameter is particularly important because it affects a compound's solubility, permeability, and distribution within biological systems, thereby playing a vital role in drug design and development. Traditional methods for estimating lipophilicity typically involve experimental techniques such as partition coefficient measurements, which are often time-consuming and resource-intensive. These methods require extensive laboratory work, the use of sophisticated equipment, and a significant amount of reagents, leading to high costs and prolonged timelines. Additionally, the accuracy of these experimental methods can be influenced by various factors, including the purity of the compounds and the precision of the experimental conditions. In response to these limitations, the field has increasingly turned to computational approaches, particularly machine learning techniques, which offer a more efficient and cost-effective alternative for predicting lipophilicity. The primary objective of this study is to enhance the prediction accuracy of lipophilicity estimates by optimizing the architecture of neural networks, a subset of machine learning models. In this research, we focus on systematically optimizing these architectural parameters to develop a neural network model that provides reliable and precise lipophilicity predictions for a wide range of organic compounds. Ultimately, the goal is to create a robust computational tool that can significantly reduce the time and resources required for lipophilicity estimation, facilitating faster and more efficient drug development processes. By leveraging the power of machine learning, we aim to advance the capabilities of computational chemistry and contribute to the development of novel organic compounds with desirable properties.