Vyasa has developed a proprietary deep learning algorithm for _de novo_ small compound design called ChemVector. ChemVector utilizes an autoencoder based neural network that can achieve >98% reconstruction accuracy on SMILES represented small compound structures along with latent space Bayesian optimization to identify and generate novel compounds that optimize variables such as logp, molecular weight and synthetic viability. ChemVector is an analytic module available in Vyasa Cortex.