Colloidal nanocrystals are miniscule pieces (< 100 nm or ∼ 1/1000th the width of a humanhair) of solid materials suspended in a liquid. Semiconductor nanocrystals have controllable optical properties like color, which make them appealing for use in displays, LEDs, solar cells and many other devices that rely on the interaction of matter with light or electricity. The color and other optical properties are related to the size and composition of the crystals, which makes control over these properties crucial to any application. Creating these crystals is often more of an art than a science, with large differences in results arising between chemicals sourced from different suppliers, laboratories, practitioners and any (even slight) changes in the process of making them. The work in this thesis is part of a broader scientific movement aimed at a more scientific understanding and rational control of the synthesis process. To keep the nanocrystals small and stable requires the use of surfactants or ligands, which are chemicals that can interact both with a charged or polar environment such as the nanocrystal and a non-polar environment such as the surrounding medium. The metal halide nanocrystals we focus on here are particularly sensitive to these ligands and the surrounding medium, as they are more similar in nature to salt, whereas most other materials used to make nanocrystals are more similar to rocks. As a result, changing the environment around the nanocrystals can transform the type, shape or composition of the crystal. The studies herein investigate the formation of metal halide nanocrystals, how transformations between them occur and how both processes are interrelated. The rapid development in precision and scale of machine learning (ML) methods capable of predictions based on large sets of data has roused the interest of scientists across many disciplines. Concurrently, improvements in the automation of chemical synthesis and charac- terization methods now allow for the generation of data on hundreds to thousands of reactions necessary to use such data science methods. Due to the complexity of chemical synthesis in general, and nanocrystal synthesis in particular, using ML to predict the outcomes of reac- tions is an attractive proposition. This thesis introduces methods from machine learning and data science to the synthesis of nanocrystals, and connects them to more physical models of this process. In this thesis, Chapter 1 outlines the important concepts related to nanocrystals, their properties, characterization methods and reactions. We then briefly introduce current trends of using machine learning and data science to understand synthesis. In Chapter 2, we use high-throughput synthesis experiments coupled to automated deconvolution of optical absorption spectra to gain an overview of the reaction landscape around one well-studied metal halide nanocrystal material, CsPbBr3. In Chapter 3, we investigate in more detail the formation and transformation processes of one material we found of central importance in the overview, the atomically thin nanosheets of OLA2PbBr4 using in-situ absorption spectroscopy and joint fits of spectral and kinetic models, as well as more traditional kinetic analysis where appropriate. Further analysis of precursor reactions is provided in Chapter 4. We then examine synthesis and properties of a family of cesium silver metal halide double perovskites in Chapter 5. Finally in Chapter 6, we discuss the implications of this work on different scientific fields.