Create and train a ML algorithm to caption images

We need a ML algorithm, f.e. built with tensorflow, which describes contents of images across all topics.

Please provide the following:

- expertise in this field (similar projects you have built)
- technology you will use
- training-dataset you will use