By controlling the variance of the radiation pressure exerted on an optically trapped microsphere in real time, we engineer temperature protocols that shortcut thermal relaxation when transferring the microsphere from one thermal equilibrium state to another. We identify the entropic footprint of such accelerated transfers and derive optimal temperature protocols that either minimize the production of entropy for a given transfer duration or accelerate the transfer for a given entropic cost as much as possible. Optimizing the trade-off yields time-entropy bounds that put speed limits on thermalization schemes. We further show how optimization expands the possibilities for accelerating Brownian thermalization down to its fundamental limits. Our approach paves the way for the design of optimized, finite-time thermodynamics for Brownian engines. It also offers a platform for investigating fundamental connections between information geometry and finite-time processes.