Pediatric cancer trials are often carried out after the adult trial has started and thus have unique features that require considerations for incorporating the adult information. In this talk, I will introduce two innovative Bayesian adaptive designs we developed for early phase pediatric cancer trials. In the first example, we propose PA-CRM, a design for pediatric phase I dose-finding trial when a concurrent adult trial is being conducted. This design uses a discounted joint likelihood to automatically and adaptively incorporate data from the concurrent adult trial into the dose assignment decision in the ongoing pediatric trial. The simulation study demonstrates that PA-CRM leads to higher chance of selecting the correct dose for the pediatric patients. In the second example, we propose BPED, a design for pediatric phase II basket trial when external data (such as adult data) is available. This design performs dual information borrowing: borrow information from the external (adult) data to the pediatric trial and between the disease subtypes within the pediatric basket trial. It also accommodates potential heterogeneous treatment effect across disease subtypes, by allowing each disease subtype belonging to the sensitive or insensitive latent subgroups. The simulation study shows that BPED yields high power to detect the treatment effect for the sensitive disease subtypes and maintains desirable type I error rate for the insensitive disease subtypes. Both designs illustrate the efficiency gain in the pediatric trial through systematic and rigorous integration of the adult data.