Heat Gun + Flour Sifter
While researching roaster designs in the early stages of my project, I came across many builds that used a heat gun for the heat source and either a dog bowl, bread maker, or flour sifter for the container and agitation method.
The heat gun + dog bowl combination is still quite manual, as it requires agitation by either the fan of the heat gun, or whisking by hand.
The heat gun + bread maker combination can be used for large batches, and provides built in agitation of the beans by leaving the bread maker on a continuous stirring setting. However, many bread makers have Teflon coatings that can deteriorate into toxic chemicals above 500°F. While the roasted beans never reach 500°F themselves, it is possible that the air coming out of the heat gun is at a much higher temperature than 500°F, so I decided against this combination for that reason.
Thus, I settled on the heat gun + flour sifter combination, which consists of a heat gun positioned underneath a flour sifter, with a drill attached to the center axle of the flour sifter, that will provide the necessary agitation for the beans. The components can be mounted on a stand, or a frame for positioning, but I chose to roast in a temporary set up, by clamping the parts to a few side tables, and positioning the heat gun on top of some bricks for proper spacing:


With this design, I was able to easily roast 3/4lbs per batch, compared to the maximum 1/4lb batch size with the popcorn popper. Not only could I roast larger batches, but more importantly I did not need to perform any action during roasting, besides monitoring temperature. Here is a side by side comparison of the two roasting methods:
This design was a great improvement for my workflow and ease of roasting. However, there were some trade-offs from the previous design. The spinning agitation method prevents the use of a temperature probe from being inserted into the bean pile while roasting, thus preventing any roasting curve data from being gathered as was possible with the air popper. Also, the device is much louder which makes it difficult to listen for audible roasting cues, such as first or second crack. This was not a problem for my goals, but for more technical projects this would have to be addressed and implemented.