Sodium-metal batteries are a promising alternative to lithium-based systems for both transportation and large-scale energy storage. They offer a cost-effective, high-energy solution due to the abundance, lower cost, and reduced toxicity of their raw materials (e.g., sodium, and cobalt-free) [1]. However, the practical viability of conventional electrolytes with sodium-metal anodes and high-voltage cathodes remains limited. Their poor electrode-electrolyte interfacial stability and low ionic conductivities lead to electrolyte decomposition, dendrite growth, and low cycling stability, ultimately reducing capacity retention and power density [2], [3]. Despite significant research efforts over the past decade, the slow discovery speed of conventional trial-and-error approaches makes it challenging to identify promising new electrolytes that meet the demands of the rapidly expanding battery market, limiting innovation and scalability. In this work, we establish an active learning workflow to accelerate the discovery of new electrolytes, applicable to both liquid and polymer electrolytes [4], [5]. We develop a high-throughput experimental platform to systematically formulate and screen new non-aqueous sodium electrolytes. We explore a chemical space larger than 1.5 x 10 9 possible compositions —comprising 11 organic solvents, 5 sodium salts, extended solvent and salt ratios, and 15 total salt concentrations. We assess the ionic conductivity and solubility of all combinations in tandem, generating the first high-quality, unified reference library of sodium-based electrolytes. The discovery speed increased by a factor of 100. For instance, in less than 1.5 months, 168 unique electrolyte formulations were obtained. By simultaneously targeting specific measurements of coulombic efficiency in high-conductivity samples, we identify several promising formulations, exceeding 11 mS cm -1 , along with high coulombic efficiency and high-voltage stability. The generated database was then used to train a machine learning model, enabling predictive selection of the next iteration of electrolyte formulations while progressively narrowing the explored chemical space. We demonstrate that by integrating our high-throughput platform with data-driven methodologies and targeted experiments, electrolyte discovery can be expedited, guiding new strategies for the design of future electrolytes for sodium-metal batteries. References [1] C. Vaalma, D. Buchholz, M. Weil, and S. Passerini, “A cost and resource analysis of sodium-ion batteries,” Nat. Rev. Mater., vol. 3, no. 4, p. 18013, Mar. 2018, doi: 10.1038/natrevmats.2018.13. [2] Y. Zhao, K. R. Adair, and X. Sun, “Recent developments and insights into the understanding of Na metal anodes for Na-metal batteries,” Energy Environ. Sci., vol. 11, no. 10, pp. 2673–2695, 2018, doi: 10.1039/C8EE01373J. [3] Y. Wang et al., “Developments and Perspectives on Emerging High-Energy-Density Sodium-Metal Batteries,” Chem, vol. 5, no. 10, pp. 2547–2570, Oct. 2019, doi: 10.1016/j.chempr.2019.05.026. [4] M. A. Stolberg et al., “A Data-Driven Platform for Automated Characterization of Polymer Electrolytes,” Dec. 23, 2024. doi: 10.26434/chemrxiv-2024-gpmm7. [5] J. Peng et al., “Human- and machine-centred designs of molecules and materials for sustainability and decarbonization,” Nat. Rev. Mater., vol. 7, no. 12, pp. 991–1009, Aug. 2022, doi: 10.1038/s41578-022-00466-5. Acknowledgments This work was supported by the Breakthrough Energy Explorer Grant. The authors also acknowledge the financial support from the MIT Postdoctoral Fellowship Program for Engineering Excellence.