Modern machine learning platforms like Tensorflow have to date been used mainly by the computer science crowd, for applications like computer vision and language understanding. But as JJ Allaire pointed out in his keynote at the RStudio conference earlier this month (embedded below), there's a wealth of applications in the data science domain that have yet to be widely explored using these techniques. This includes things like time series forecasting, logistic regression, latent variable models, and censored data analysis (including survival analysis and failure data analysis).
The keras package for R provides a flexible, high-level interface for specifying machine learning models.(RStudio also provides some nice features when using the package, including a dynamically-updated convergence chart to show progress.) Networks defined with keras are flexible enough to specify models for data science applications, that can then be optimized using frameworks like Tensorflow (as opposed to traditional maximum-likelihood techniques), without limitations on data set size and with the ability to apply modern computational hardware.
For learning materials, RStudio's Tensorflow Gallery provides a good place to get started with several worked examples using real-world data. The book Deep Learning with R (Chollet and Allaire) provides even more worked examples translated from the original Python. If you want to dive into the mathematical underpinnings, the book Deep Learning (Goodfellow et al) provides the details there.
RStudio blog: TensorFlow for R