A blackbox optimization problem is such that its objective(s) and constraints are provided by a computer code or an experiment, a simulation seen as a blackbox. This blackbox may fail to evaluate, be time-consuming, be contaminated by noise, etc. Most importantly, analytical expressions of the problem, including derivatives, are unavailable. In such a context, optimization methods that do not rely on derivatives are needed. These derivative-free optimization algorithms exist since the 50’s, but have rapidly evolved over the last 25 years. This presentation introduces several examples of applications, including hyperparameter tuning and the design of a solar power plant, and gives an overview of the different families of methods, with a focus on algorithms that are not heuristics since they possess mathematical guarantees of convergence. Several features will be highlighted, such as constraints handling, multiobjective optimization, discrete variables as well as the use of models and surrogates. The NOMAD software package, designed for blackbox optimization, will be used to illustrate these features.