This study compares engineering expert problem-solving on a highly constrained routine problem and an ill-defined complex problem. The participants (n=7) were recruited from two large public Research I institutions. Using a think aloud methodology, the experts solved both routine and non-routine problems. The protocols were transcribed and coded in Atlas ti. The first round of coding followed a grounded theory methodology, yielding interesting findings. Unprompted, the experts revealed a strong belief that the ill-defined problems are developmentally appropriate for PhD students while routine problems are more appropriate for undergraduate students. Additional rounds of coding were informed by previous problem solving studies in math and engineering. In general, this study confirmed the 5 Step Problem Solving Method used in previous challenged based instruction studies. There were observed differences based on problem type and background knowledge. The routine problem was more automatic and took significantly less time. The experts with higher amounts of background knowledge and experience were more likely to categorize the problems. The level of background knowledge was most apparent in the steps between conducting an overall energy balance and writing more problem specific relationships between the variables. These results are discussed in terms of their implications for improving undergraduate engineering education. / text