This study explores the potential of machine learning to accelerate the design of materials with tailored nonlinear optical properties. Specifically, we develop a machine learning model to predict the first and second hyperpolarizabilities of molecules, which are crucial parameters governing a material’s nonlinear optical response. Using a subset of 100 molecules from the QM9 dataset, we train a model with 0.2 test size on molecular features such as band gap, dipole moment, polarizability, and molecular energy. The R2 score of First hyperpolarizability and second hyperpolarizability are 0.9199 and 0.9414 respectively. The mean absolute error of first and second hyperpolarizabilities are 28.332 and 80.008 respectively. This work underscores the promise of machine learning as a computationally efficient alternative to traditional Density Functional Theory (DFT) calculations for screening and designing materials with enhanced nonlinear optical properties. The ability to accurately predict these properties has significant implications for the development of high-performance photonic devices, particularly electro-optic modulators and switches crucial for high-speed data transmission in optical telecommunications. Future research will focus on improving the model’s accuracy for first hyperpolarizability prediction and expanding the feature set to encompass a wider range of molecular characteristics, ultimately contributing to the advancement of next-generation photonic technologies.