Electrifying the transportation sector will be critical to building a sustainable future. Successfully meeting mobility needs with electrical energy, however, will require careful planning and coordination between drivers, grid operators, and policy-makers. Mass electrification, if orchestrated naively, will lead to tremendous increases in both power and energy consumption and will disrupt power grid operations. This thesis presents a set of methodologies for ensuring that transportation can be electrified at scale in a way which respects both grid constraints and driver mobility needs. Ultimately, the objective of this body of work is to help cultivate a symbiotic relationship between the grid and the vehicles it will fuel. The first section of this thesis is focused on the challenge of facilitating coordination between drivers and grid operators. Specifically, this chapter presents a novel mechanism for EV load scheduling in which drivers allow their defferable EV charging loads to be controlled by a central planner in exchange for a monetary incentive. This mechanism enables grid operators to harness EV load flexibility in order to shape charging loads to meet a wide variety of objectives, e.g., minimization of peak loads, load variance, or renewable curtailment. To illustrate the practical efficacy of the proposed mechanism, Chapter 2 presents results from a smart charging pilot program. The pilot, spanning seventeen months and including thirty-five participants, found that the proposed mechanism was highly effective in shifting EV charging loads. Participants made their charging loads flexible during over half of all charging sessions and provided an average of eight hours of flexibility. Furthermore, driver participation remained steady throughout the course of the pilot, suggesting that the proposed mechanism may be a viable approach to load control long-term. The following chapter deals with another key challenge: planning for electrification underuncertainty. We address this challenge through chance-constrained optimization, a branch of optimization involving problems whose constraints are required to hold with high probability with respect to the distribution of an uncertain variable. Chance constrained optimization provide a framework for planning infrastructural or energy needs under various sources of uncertainty such as renewable generation, grid load, or mobility needs. In the first portion of Chapter 3, we present a non-parametric approach to the approximation of solutions to chance constrained problems. Our approach gives rise to tractable optimization problems which are guaranteed to generate a feasible solution to the initial chance constrained problem with high probability given access to a finite number of samples of the random parameter. In the second portion of this Chapter, we present a planning problem of practical interest which admits a formulation as a chance constrained problem. Specifically, we consider the problem of planning battery capacity in a setting with uncertain mobility needs. We consider two potential scenarios, one in which battery resources are private and one in which drivers are able to share resources amongst themselves. Using real-world mobility data to conduct an empirical study, we find that resource sharing can substantially reduce the amount of battery capacity required to meet commuting needs with high probability. The methods discussed in this thesis rely critically on the availability of real-world data. In the final chapter, we consider the problem of optimally acquiring such data. Specifically, the final chapter deals with the task of designing data collection in order to learn about mobility patterns with maximum efficiency. Motivated by the desire to model vehicle duty cycle behavior, we formulate the data collection task as a Bayesian optimal experiment design problem. Though intractable to solve exactly, the problem admits a convex relaxation which we solve to generate approximate solutions specifying an near-optimal experiment design. We evaluate the quality of the resultant experiment designs by simulating data collection using vehicle mobility data from the FleetDNA database, and find that BOED-based experiment design leads to faster reduction of error covariance than alternative designs.