Modeling biomolecular machines with the probability flows on Markov networks often entails computing node probabilities and edge fluxes with Master Equations and correlating machine performance with entropy production. But this alone is not sufficient for design, optimization and evolution because it doesn’t treat cost-benefit tradeoffs, or topological misflows (backsteps, futile cycles, ineffective actions), or differential properties for flow design. Here we develop CFT Design, based on the recently developed Caliber Force Theory (CFT). We apply it to: designing faster molecular motors through "traffic control"; optimizing speed, energy, and accuracy in kinetic proofreaders simultaneously to break their trade-offs; and designing better enzyme inhibitors based on their network topologies. CFT Design provides a general framework for optimizing nonequilibrium flows.