Hey, I am glad that you liked the article. Basically TensorFlow and MLKit are two different building blocks and yes you can use them independently. In this article, I have used TensorFlow for building the model and then, later on, to integrate that model with Android app I have used MLKit, which provides ease of local/remote model deployment. In the current article, a model was deployed inside the android app but it could have been done another way as well like deploying it on Google Cloud and use MLKit API.

I hope I clarified your doubt. In case of any queries please feel free to post.