Interview with Pulkit Verma: Towards safe and reliable behavior of AI agents