Interesting article here - probably just an excuse to make their coffee habits tax deductible!
I think that its a damned shame that caffeine is invariably the subject of chemical investigations. These scientists could just as easily use spectroscopy to analyse how levels of different types of sugars and caramels change throughout the roast, which might actually produce some useful research.
My understanding is that caffeine basically doesnt really do much as the beans roast; the increase in caffeine percentage is due to a loss of other mass (mainly water), but the total mass of caffeine doesnt really change much. If that is the case, and it has already been proven, then this project would be an especially stupendous waste of time. It basically wouldnt tell us anything that we couldnt already work out with a decent set of scales (costing far less than an IR spectrophotometer, I might add).
Now, it looks like they are getting a whole bunch of data in the 1100 to 2200 nm band, so it might be that the article is just dramatically selling short the scope of the project. It has been a while since I have done any IR spectroscopy, so I cant remember off the top of my head if anything interesting lies between those wavelengths.