Is use of unconventional design elements or layout something frowned upon in academic context?
Good design is invisible. The goal of design is to increase understanding/clarity. If your design is truly good then it will go unnoticed but your content will be better understood. If your design is noticed and distracts, then it is bad. I feel these principles are universal.
So to answer your question, yes, good design is worth it (it increases the amount that your content is understood), but just adding "design" elements without a good understanding of the audience and their expectations will not likely result in good design.
It all comes down to a cost/benefit analysis. But, there is little risk to improve the design, graphics and typography of your documents (theses, figures, presentations). There is little risk that it backfires if you present a higher-quality document. In fact, the only case I can think of is if it seems that form took over content: i.e., if you have a very shiny designed presentation with just-meh scientific content, the contrast might draw attention.
One thing that might be a problem is if you put too much theatrics, 3D effects, animations, cartoons… I had a colleague who used every single “animation” possible (it was the early days of Apple's Keynote and its nice 3D effects) in the same presentation, and it was simply too much. It distracted people from his message.
Finally, coming back to the cost/benefit analysis: I believe that as in everything, 20% of the work can get you 80% of the reward if you choose wisely. People will have different pet peeves, but the areas which I think you should polish for presentation slides are:
- Graphics quality: no pixelated crap
- Consistency between graphics and text, and self-consistency of graphics: same quantities reported and plotted, same units, consistency between graph scales (as much as possible), etc. Sometimes you take pictures from an earlier paper, and they don't quite match what you are showing with them. Avoid things like “graph on the left is concentration, graph on the right is volume fraction” when they can be converted straightforwardly.
- Careful about background colors: to me, this makes the different between decent slides and good slides (for the presentation, not for the scientific content). If you use a colored background (not saying you should), don't include graphics with white background. Try to use graphics with transparent background (easy with vector graphics, use PNG with alpha channel for bitmap images).
- If visualization requires it, use movies to show a complex system: time evolution of spatial distribution, autorotation of a structure you present if it makes it clearer, etc.
There are different kind of academic documents, and it might change from one field to another. For instance, journal/conference papers often have a required style, and so there is little room for improving the overall design of the paper.
About the CV, I have seen many places asking for a specific style for the CV (i.e., they give you a Word document to fill in ...), but when they don't, as long as the content respects the traditional Education/Experience/Publications, then it shouldn't be a problem.
When it comes to presentations, I would say it's a bit trickier. The overall impression I've had when talking with colleagues at a conference, is that the quality of the speaker matters first. I've attended excellent presentations where the slides where black and white powerpoint, because the speakers were merely using them as a support to put the keywords and the important formulae. Conversely, I've attended boring presentations where the slides where very nice, just because the speaker was not comfortable speaking.
In other words, I don't think putting time and effort in the design of slides will backfire, and it will not necessarily bring you bonus points, what matters is the quality of the presentation itself.