On-demand aviation refers to an envisaged air taxi service, using small, autonomous, vertical-takeoff-and-landing, battery-powered electric aircraft. A conceptual design and optimization tool for on-demand aviation is presented in this thesis. The tool uses Geometric Programming, a class of optimization problems with extremely fast solve times and for which global optimality is guaranteed. The optimization model consists of a vehicle, a sizing mission, a revenue-generating mission, and a deadhead (non-passenger-carrying) mission. Cost per trip, including the additional cost due to the deadhead mission, is used as the objective function. Vehicle noise is computed during post-processing using a semi-empirical method. The tool is used to conduct a study of on-demand aviation from a vehicle design perspective.A trade study is conducted between several different on-demand aircraft configurations. Four configurations are viable: the lift + cruise configuration, the compound helicopter, the tilt wing, and the tilt rotor. Configurations with a higher lift-to-drag ratio, but a higher disk loading, generally weigh less and cost less to operate; configurations with a lower lift-to-drag ratio, but a lower disk loading, are quieter. Using New York City as an example market, it is shown that an on-demand air service will cost significantly less as compared to current helicopter air taxi operations. The two most important costs are pilot salary and battery amortization. If these two costs can be reduced (via vehicle automation and reduced battery manufacturing costs respectively), an on-demand air service becomes competitive with current car ridesharing on the basis of cost per seat mile. Therefore, on-demand aviation has the potential to become a system for everyday commutes.Technological assumptions and vehicle requirements, especially mission range, battery energy density, vehicle autonomy level, battery manufacturing cost, and reserve requirements, have significant impacts on vehicle weight and cost. Vehicle noise can be reduced through the careful selection of key design parameters. However, envisaged noise requirements cannot easily be met, even with the most generous long-termThe author wishes to thank all those who contributed their insights to this work. In particular,