This paper introduces the fabrication and calibration of a soft shape memory alloy actuator with an integrated liquid metal sensor. The actuator is capable of transforming from an unactuated soft curled shape to an actuated rigid straight shape when it is electrically activated. The surface-bonded sensor is a capacitive strain gauge capable of tracking actuator curvature, which is composed of microfluidic channels of liquid metal alloy embedded in a soft silicone elastomer. The sensor has limited impact on the mechanical properties of the actuator due to being soft and lightweight. The fabrication procedure of the actuator is demonstrated in detail, and it is explained how the actuator can be easily fabricated using rapid prototyping techniques, such as laser cutting and stencil lithography. Then, the calibration procedure shows how the capacitance of the bonded strain gauge can be related to the actuator curvature by aligning capacitance data from the sensor with the actuator’s curvature captured by a fast camera. Finally, we implemented a closed-loop control strategy to show the effectiveness of the integrated sensor in improving the actuator performance. The used control scheme provides a method for optimizing actuation in a way that maximizes actuation amplitude. For optimal control, we use a learning-based approach with a covariant matrix adaptive evolutionary strategy (CMA-ES). It is shown that small change in either applied voltage or actuation time will lead to a large difference on the actuation performance.