The new infrastructure bill promises to invest in repairing and replacing the aging critical infrastructure in the US, particularly roads and bridges. Ultrahigh-performance concrete (UHPC) is a "robust solution for highway infrastructure" since the high compressive strength (>120 MPa), high tensile strength (>5 MPa), homogeneity, and the superior durability of UHPC make it highly reliable wherever it is used (Du et al. 2021). This research aimed to produce and evaluate the performance of low-cost UHPC mixtures using local ingredients. To that end, in the first section of this dissertation, a database consisting of 139 mixture proportions of non-proprietary UHPC formulations, and their workability (flow) and compressive strength data was developed. This database was expanded using prior data of 83 UHPC mixtures developed at Penn State. Random forest-based machine learning (ML) models were developed using this data to first predict the mixtures' flow and compressive strength. Further, physics-based knowledge of the impact of particle size (of aggregates and cementitious materials), particle packing, and chemical composition of the cement was incorporated into the model. This endeavor improved the generalizability of prediction models. In the second part of the dissertation, 130 UHP mortar and concrete mixtures were developed using local ingredients and were evaluated. A plan was devised and executed to determine the impact ingredient properties, mixture proportions, and the temperature of fresh concrete on the properties of UHPC. This experimentation also allowed insight into the robustness of the developed UHPCs to quantify their sensitivity to variability in the mixture constituents and proportions. This research study determined lower-cost alternatives for expensive mixture constituents like quartz filler and quartz sand. At the end of this section, several low-cost UHPC mixtures were developed that met the requirements of FHWA and AASHTO guidelines. Reliably characterizing the mechanical performance, especially the tensile-stress-strain behavior of UHPC mixtures, is critical to determine whether the UHPC exhibited strain-hardening behavior. However, measuring this response by applying uniform loading in tension without pre-maturely cracking the test specimens or inducing large bending strains in the process is challenging. Therefore, in the third section of the dissertation, a UHPC tensile testing setup in compliance with the AASHTO T397-22 test method was built. This set-up involved equipping an 11-kips load cell with custom-made flat grips fabricated following AASHTO recommendations. In addition, an aluminum extensometer was fabricated to measure the average strain in the gauge section, and a data acquisition system was set up to accurately record the load and displacement signals. As a result, the mixtures' extent of strain hardening, and pre-crack localization strain capacity were consistently determined. This setup further allows for the optimization of the fiber reinforcement to reduce the cost of UHPC. Finally, a comprehensive literature review was carried out with the objective of assessing the risk of, and formulating strategies to mitigate, shrinkage cracking in UHPC mixtures. UHPC mixtures tend to be dominated by cementitious materials, composing greater than 60% of the volume of concrete. A high content of cementitious materials and low water-to-cementitious material (w/cm) ratio render UHPC mixtures vulnerable to shrinkage cracking and damage, particularly in connections between prefabricated bridge elements. Thus, the last part of the dissertation reviewed the risk of shrinkage-based cracking in UHPC, the role of fiber reinforcement, and finally assessed the effectiveness of mitigation measures. The outcome of this dissertation is the reduction of the cost of non-proprietary UHPC that were compliant with and exceeded FHWA and AASHTO design guideline performance. It also adds to the know-how in the reliable production of low-cost non-proprietary UHPC mixtures by offering consistently produced and collected data. ML modeling offers efficient and robust tools for future UHPC producers to optimize UHPC mix designs with minimal trial and error.