Over the course of the COVID-19 pandemic, governing bodies and individuals have relied on a variety of non-pharmaceutical interventions (NPIs) to control the transmission of SARS-CoV-2, which posed an acute threat to individuals' well-being and consistently impacted economic activities in many countries worldwide . NPIs have been implemented at varying levels of severity and in response to widely-divergent perspectives of risk tolerance . Now, concurrently with the introduction of multiple SARS-CoV-2 vaccines, the world looks optimistically to a``return to normality".This work was supported by the AB Nexus seed grant from the University of Colorado . In this work, we propose a multi-disciplinary approach, combining transmission modeling with control and optimization theory, to examine how risk tolerance and vaccination rates will impact the safe return to normal behavior over the next few months . To this end, we consider a version of the Susceptible-Exposed-Infected-Recovered transmission model that accounts for hospitalizations, vaccinations, and loss of immunity . We then propose a novel control approach to calibrate the necessary level of NPIs at various geographical levels to guarantee that the number of hospitalizations does not exceed a given risk tolerance (i.e., a maximum allowable threshold). Our model and control objectives are calibrated and tailored for the state of Colorado, USA . Our results suggest that: (i) increasing risk tolerance can decrease the number of days required to discontinue all NPIs; (ii) increasing risk tolerance inherently increases COVID-19 deaths even in the context of vaccination; (iii) if the vaccination uptake in the population is 70% or less, then return to normal behavior within the next year remains risky . Furthermore, by using a multi-region model accounting for travel, our simulations predict that: (iv) relaxation should take into account regional heterogeneity in transmission and travel; and (v) premature relaxation of NPIs, even if restricted only to low-density regions, will lead to exceeding hospitalization limits even when highly-populated regions implement full-closures . Although the simulations are performed for the state of Colorado, the proposed model of transmission and control methods are applicable to any area worldwide and can be utilized at any geographical granularity.