The shift towards a more sustainable energy economy is one of the imperative challenges facing humanity today, and balancing prosperity against the risks of irrevocable climate change will require policy adjustments and scientific innovations on a global scale. In particular, it is essential to move away from burning fossil fuels to meet our energy needs; rising atmospheric CO2 has already contributed to ocean acidification and record high temperatures, and the dangers only increase with every ton of CO2 emitted. Fortunately, wind and solar radiation provide vast resources for renewable energy, and remarkable progress has been made in the past several years towards incorporating these sources. As the use of renewable energy generation rises, so too does the need for efficient energy storage and conversion that are not predicated on the use of fossil fuels. Electrochemistry offers one piece of the solution through fuel cells, batteries, and other technologies. The drive to discover and refine catalysts for these electrochemical reactions is therefore of critical importance to our shared sustainable energy future. Catalyst design has benefited from the close integration of experiment and theory in a cyclical framework whereby new materials are synthesized, characterized, tested for electrochemical performance, and used to improve predictions for future catalysts. A similar framework is used in this dissertation as we delve into each part of the catalyst development cycle. We begin with materials synthesis of nanoparticles, which are of scientific interest for their unique properties compared to bulk materials. Inert gas condensation is introduced as a method for nanoparticle synthesis, and we present several systems including NiFe, Mn oxides, and other transition metals. We observe several unusual morphologies, including cubic particles and the alignment of particles on surface defects. In addition, we study catalytic activity for the oxygen evolution reaction (OER) on both NiFe of varying sizes and Mn oxide promoted with Au. We demonstrate that inert gas condensation is a highly versatile method for synthesizing nanoparticles both for fundamental studies and as electrochemical catalysts. We then focus on the details of one specific catalyst: CuAg for the oxygen reduction reaction (ORR). The ORR is a key component of fuel cells and metal-air batteries, and developing efficient and cost-effective catalysts for this reaction will entail improving our understanding of catalyst activity. We find that CuAg nanoparticles outperform either Cu or Ag nanoparticles, and that they are on par with thin films of similar compositions. To elucidate the origin of this heightened activity we use a combination of density functional theory (DFT) and in situ characterization. X-ray absorption spectroscopy (XAS) allows us to follow the electronic state of our catalyst under reaction conditions, and while we see little change in the electronic or geometric state of the Ag atoms in CuAg, the Cu atoms in CuAg are markedly different than in pure Cu. DFT predicted that Cu atoms in a Ag lattice would have dramatically different d-band states and a smaller oxygen binding energy, and our in situ experiments confirmed that Cu atoms in CuAg are more reduced than in Cu at ORR-relevant potentials. CuAg is revealed to owe its enhanced activity not to a small change in Ag, the more active metal alone, but to a substantial modification of Cu that boosts the overall performance. We hope that better understanding this system will contribute to the design of highly active non-precious catalysts for the ORR. Traditionally new catalysts for a reaction are chosen based on a combination of conventional theory calculations such as DFT and educated guesswork informed by scientific insight. However the vast search space of possible catalyst materials and the wealth of computational and experimental data for reactions studied over decades opens the possibility to use machine learning to speed the iterative design process. In the final portion of this work we consider the application of machine learning to case studies in both computational and experimental materials science. To start, we examine several algorithms for predicting metallic glasses on ternary alloys from a historical dataset based on their compositions alone. Using the two best models, we then investigate combining sparse historical data with new high-throughput data and find that more data is not always better. On the other hand, materials science encompasses many questions for which the data is much less plentiful. One strategy to maximize the value of small datasets is transfer learning, in which the outputs of one model inform subsequent models. We apply transfer learning to experimental Ni superalloy mechanical properties and nitric oxide reduction reaction computational data, and we determine that in both cases transfer learning is an effective way to improve model accuracy without collecting new data. In summary, this dissertation explores each step of the catalyst development cycle, from nanoparticle synthesis, to electrochemical testing, advanced in situ characterization, and predicting new materials via machine learning. This work aims to present fundamental insights on catalytic activity as well as several avenues for future catalyst development with the goal of contributing to a more efficient energy future.