Next-generation mobile core networks are being designed to support a variety of latency sensitive applications based on emerging virtual, augmented or mixed reality technologies. A cloud-native approach for 5G core has been proposed to meet the diverse service requirements of NextG while reducing both CAPEX and OPEX. In this context, microservice architecture for network function virtualization is generally considered to be suitable for meeting NextG service requirements. Despite many advantages, the cloud-native core raises new challenges in the design of NextG systems for latency critical applications. An approach to achieving diverse QoS requirements is proposed in this paper. Specifically, the design is based on an orchestrator called the MEC-Intelligent Agent (MEC-IA) which enables dynamic compute resource distribution and network slice assignment in the core for improved QoS. The MEC-IA framework realizes resource management by intelligently assigning UEs to the access and mobility management function (AMF) while also performing slice provisioning. Simulation results are presented for the proposed MEC-IA framework showing the median control plane delay reduced by a factor of 1.67 ×. Further, robustness of the system improves significantly, reflecting a better overall user experience since the percentage connection dropped at 3 × traffic volume reduces by 1.5 × and slices assignment increases by 1.4 × across all slices, even when the traffic arrival is skewed.