The next generation of cosmological experiments will significantly strengthen constraints on two of the most interesting inflationary observables, the tensor-to-scalar ratio r and the non-Gaussian component of the primordial scalar fluctuation fNL. I describe why these quantities are very sensitive probes of high energy physics, and how inflationary models which produce measurable levels generally must be embedded in a complete UV framework like string theory to be trustworthy. I then discuss whether and how inflationary models in string theory can produce measurable r and fNL.