The development of automated experimental facilities and the growing trend of experimental data digitization brought enormous opportunities for radically advancing laboratories. As many laboratory research tasks involve predicting and understanding previously unknown physical or chemical relationships, the availability of experimental data enables machine learning (ML) approaches to substantially accelerate the conventional design-build-test-learn process. In this talk, we will introduce software packages, including RobustGaSP, RobustCalibration, FastGaSP, DDM-UQ, and AIUQ, for constructing scalable surrogate models to predict complex relationships, enabling efficient inverse estimation and materials characterization. Furthermore, we introduce how artificial intelligence (AI) agents based on large language models can help researchers learn background knowledge in materials science and data science to accelerate discovery processes. We will introduce various case studies enabled by these tools, including universal phase identification of block copolymers by small-angle X-ray scattering, ab initio uncertainty quantification in scattering analysis of microscopy, fast phase prediction of charged polymer blends by white-box machine learning surrogates, and reliable emulation of classical functional theory calculation by active learning with error control.