Stan (and hence brms) will provide full Bayesian estimation for all your parameters. In most formulations, the varying intercepts are assumed to have zero mean, but yours appear to be fitted far away from zero. Hard to say what is the problem without examining the exact code and data that produced this, but a possible cause is that you have removed the overall intercept? This most likely suggests a problem with your model, but once again hard to be specific what exactly is the problem.
It feels like you might benefit from going through same course on the basics of Bayesian inference (see e.g. Understanding basics of Bayesian statistics and modelling). A very brief response is: 1) you usually don’t want just a point estimate. 2) Stan (and other MCMC algorithms) will give you samples from the posterior, which then let you compute expectations of various functions with respect to the posterior distribution (e.g. expectation of the identity function is the mean), which expectations (and hence which point estimates and measures of uncertainty) are most relevant to you applicaiton depends on the question you are asking. 3) In most cases, posterior mean and some uncertainty interval (e.g. 95% central credible interval) is a good start.
Best of luck with your model!