In Oncology, increasing the treatment dose generally increases both efficacy and toxicity. With continuous dose, the goal is to analyze an existing dataset to estimate an optimal dose for each (future) patient based on their clinical features and biomarkers. In this paper, we propose an optimal individualized dose finding rule by maximizing utility functions for each patient and satisfying the average toxicity tolerance. This approach maximizes overall efficacy at a prespecified constraint on overall toxicity. We model the outcomes using logistic regression with dose, biomarkers and their interactions. To incorporate the large number of biomarkers and their interactions with dose, we employ the LASSO with linear constraints on the dose related coefficients to constrain the dose effect to be non-negative. Simulation studies show that this approach can improve efficacy without increasing toxicity relative to fixed dosing. Constraining each patient estimated dose-efficacy and dose-toxicity curves to be non-decreasing improved performance relative to standard LASSO. The proposed methods are illustrated using a dataset of patients with lung cancer treated with radiation therapy.