We use machine learning (ML) techniques to construct models for various databases/problems related to chemistry, biology, and physics. Machine learning is used in molecular reaction dynamics for systems consisting of only a few atoms to protein-ligand/DNA/RNA binding, and local structure determination in solid/liquid-like systems consisting of a few hundred to thousands of atoms or particles. Multidimensional potential energy surfaces for polyatomic molecules are constructed using Gaussian process regression (GPR), kernel ridge regression (KRR), and artificial neural networks (ANN) methods for studying reaction dynamics or spectroscopy. ANN based models are generated to predict different reaction observables e.g., reaction probabilities, cross sections, rates, and distributions of products.  Binding between protein and DNA/RNA or ligands are being studied using different supervised ML algorithms (Random forest, support vector machine, ANN, convolutional neural network (CNN)). ML based models  (Random forest, Support vector machines) are constructed to diagnosis various diseases using mass spectra obtained from affected tissues or blood samples. Characterization of different phases in the heterogeneous medium is performed using dimensionality reduction techniques (Autoencoders, Principal component analysis, t-SNE) and unsupervised ML based clustering algorithms (Agglomerative clustering, DBSCAN) and Gaussian mixtures model.