Hello, I am implementing recommendation recommendation system for long tail web service using python, tensor flow and keras.

I need help like which dataset I need to use and any more help would be grateful.

I don't demand any code but if any one can recommend me the road map of the approach, it would be better. My passionate areas are deep learning, machine learning and python programming.

I know that long tail is widely by giant companies like google and many others but I need to understand what should be my approach to make such a recommendation system.

Since this is your 'passionate area', you really need to learn something about what you're asking about first:
https://en.wikipedia.org/wiki/Long_tail

A recommendation system requires something to recommend. You don't just build it stand-alone. Long tail is just a term for how google and amazon, and others do marketing of both popular items and niche market items. The system is really just one application of artificial intelligence. No one is going to tell you how to build such a system, because the people who know how to do it don't want everyone else to know. It's worth billions of dollars to keep it private.