The air flow velocity field in an internal combustion engine is fundamentally involved in all aspects of the combustion process, broadly affecting engine performance, including fuel economy, stability, heat release rates, and exhaust emission. Unfortunately, common and advanced methods used to design engine surface geometries which control this velocity field typically rely on simple non-reacting models which fail to reflect true complex combustion behaviors. An alternative but presently undeveloped approach which could overcome this challenge, is to integrate shape optimization, additive manufacturing, and firing engine experiments in a process which creates designs driven directly by engine performance measurements and targets. In this work, an initial experimental approach to the shape optimization of a tumble-inducing orifice within an internal combustion engine intake runner was developed and demonstrated. The process developed here was based on a basic genetic algorithm, where in each generation, a set of intake runner designs were 3-D printed and the tumble intensity generated by each design was measured using particle imaging velocimetry. Design convergence was achieved for three free design parameters after 16 experiments, likely far less than the number required for more traditional design-of-experiment optimization approaches. The results indicated that an elitist strategy, where the best gene is retained in the next generation, was critical to manage run-to-run variability. Major challenges identified in the trial of this method were a significant time bottleneck for 3-D printing and the negative effects of run-to-run variability on convergence detection. Promising future investigations could include the use of shape morphing surfaces and efforts to combine experimental and CFD-based design optimization methods.