What do you like best?
The API is straightforward, well documented, and makes it easy to prototype and test models with a minimum of outside setup. They also offer a fairly large range of instance images which cover every deep learning framework that I have used and some that I have not. The team also has been extremely responsive to concerns, and fixed several major pain points when they were brought up. This makes me confident in both the team and the future of the product .
What do you dislike?
I don't think it's an efficient alternative yet for large projects- it's great for experimentation, testing, and small-to-medium size projects. If you are doing research, experimentation, or something along those lines it is perfect.
Recommendations to others considering the product
If you are looking for the flexibility, ease of setup and use, then Floydhub is a great alternative to spinning up an instance. The runway for me from learning about the software to running my first job was less than an hour.
What business problems are you solving with the product? What benefits have you realized?
It makes it easy to test, prototype, and explore modelling options.