Biarticular actuators inspired by human muscles offer significant potential for enhancing robotic legged locomotion, but their effectiveness critically depends on geometric design and actuator placement. This paper presents a geometry optimization framework for a bioinspired robotic leg combining biarticular and monoarticular actuators. Using human motion capture and joint torque data, we formulate a constrained optimization problem to maximize actuator torque feasibility by optimizing leg geometry. The optimization is subject to constraints that maintain biarticular actuator moment arm ratios close to biomechanically optimal values during walking and running. A multifidelity optimization approach efficiently navigates the complex nonlinear design space under actuator and mechanical constraints. Simulation results show that the optimized geometry preserves favorable moment arm ratios (1.5–2.5) throughout thousands of running and walking gait cycles, including balance recovery scenarios, yielding high safety margins and improved mechanical performance. These findings highlight the essential role of geometry optimization in achieving biomechanically realistic and efficient biarticular actuation for dynamic legged robots.