Struggling to design a scalable microservices architecture with Kafka – where to draw service boundaries?

I'm kicking off a new project and opting for a microservices approach, with Kafka as the backbone for inter-service communication. My biggest hurdle right now is defining clear, independent service boundaries without ending up with either a monolith or a 'nanoservice' mess. What are the key heuristics or design patterns the community uses to decide where one service ends and another begins, especially when dealing with shared data models or complex business processes? Looking for practical advice beyond textbook definitions!