All the things you mentioned are achievable. Some of the things require a fair amount of integrating with external APIs and tooling:
1> Once a user logs in welcome the user and start interacting with the user through voice assistance.
This should be fairly straight forward with text. However you will then need some sort of text to speech to get the voice.
2> To help a user search for a product and its recommendations
Can be achieved by having the assistant make function calls to an API that does the actual searching and recommendation.
3> Ti uploads an image and searches for similar kinds of products
Similar to the point above but will need to use sort of function call that can do the actual image search OR use a model that can extract details of what is in an image and then make a function call to search for recommendations.
4> Help to resolve any F&Q
Knowledge retrieval from files uploaded to the assistant could handle this.