The agent technology is used to develop systems that perform several complex tasks. This thesis presents the MAS-School method for modeling and implementing intelligent agent-based systems. The method presents a systematic approach to support a disciplined introduction of machine learning techniques in multi-agent systems from an early stage of design. The proposed approach encompasses guidelines to both the design and implementation phases of an agent-based system. It is based on an incremental development strategy that largely relies on simulation and testing techniques. This thesis also presents the ASYNC framework that is composed of software engineering tools for building agent based system for asynchronous, cooperative and intelligent agents. This thesis presents four complex applications that used the proposed method and framework in the design and implementation phase. The first case study presents an application that discovers the most appealing offerings for consumers in a retail market. The second case study presents a virtual marketplace for buying and selling goods with automatic negotiation. The third case study is a multi-agent system for a complex procurement scenario with interdependent and simultaneous auctions. This system achieved the third place in the 2004 TAC Classic competition. The fourth case study is a multi-agent system for a PC manufacturer scenario based on sourcing of components, manufacturing of PC's and sales to customers.