LEA (Learning Environments Agent) is a web-based software system for advanced multimedia and virtual-reality education and training. LEA consists of three fully integrated components: (1) unstructured knowledge-base engine for lecture delivery; (2) structured hierarchical process knowledge-base engine for step-by-step process training; and (3) hierarchical rule-based expert system for natural-language understanding. In addition, LEA interfaces with components which provide the following capabilities: 3D near photo-realistic interactive virtual environments; 2D animated multimedia; near-natural synthesized text-to-speech, speech recognition, near-photorealistic animated virtual humans to act as instructors and assistants; and socket-based network communication. LEA provides the following education and training functions: multimedia lecture delivery; virtual-reality based step-by-step process training; and testing capability. LEA can deliver compelling multimedia lectures and content in science fields (such as engineering, physics, math, and chemistry) that include synchronized: animated 2D and 3D graphics, speech, and written/highlighted text. In addition, it can be used to deliver step-by-step process training in a compelling near-photorealistic 3D virtual environment. In this paper the LEA system is presented along with typical educational and training applications.