Of course!

You will need to store your trained model in a persistent store that your Heroku app has access to. (can be storage like S3, or a persistent database)

You just need to download your model before your application starts :)

If you want to use TensorFlow-Serving, you can build your own Docker image that contains your code + TensorFlow-Serving executable.

(https://devcenter.heroku.com/articles/build-docker-images-heroku-yml)