There is a growing trend to incorporate the Adaptive Automation (AA) paradigm into the pilot – aircraft environment which has led to contrasting results. On one hand, this has brought about a reduction of pilot workload in addressing mundane tasks; on the other hand, it has raised a number of concerns related to issues such as reduced pilot attention and situational awareness, over-reliance on automation as well as dominance (pilot vs. automation). With the wide acceptance of automation in other sectors of society, it is reasonable to assume that Adaptive Automation is going to be the norm in this context as well. Thus, maximization of automation will depend on the degree to which the “pilot-automation” system is able to optimize task allocation and control between the human operator and the automation. One approach to achieve such a balance is to characterize human workload levels – both physical and mental – to create indicators and thresholds of when automation is most useful. In this effort, we aim to generate pilot performance profiles by applying a number of unsupervised machine learning algorithms to flight data from flights of a remote controlled fixed wing aircraft. By classifying flight data in an unsupervised manner, we hypothesize that 1) the workload levels of the human operator will be reflected in the recorded flight data, and 2) unsupervised learning methodologies can classify flight data into identifiably different performance profiles. This research will explore how using different parameters and features from flight data and clustering algorithms (for example, k-means, hierarchical trees, neural networks, and hidden Markov models) affect the generation and definition of performance profiles. The results could eventually be used to achieve a tighter integration in the “pilot-automation” system.