This review discusses recent developments in turbulence modeling, emphasizing the integration of data-driven approaches with traditional turbulence modeling techniques. Turbulence modeling has traditionally relied on Reynolds-averaged Navier-Stokes (RANS) equations, which require assumptions to close the system of equations. However, these assumptions can introduce inaccuracies, limiting the predictive ability of RANS models. Recent advances in data-driven modeling, statistical inference, and machine learning offer new ways to improve turbulence models by systematically incorporating data to quantify and reduce model uncertainties. The review outlines the four layers of simplifications in RANS closures and discusses the challenges in assessing the predictive nature of turbulence models. It highlights the importance of uncertainty quantification (UQ) in evaluating the reliability of model predictions. UQ aims to rigorously measure and rank the effect of uncertainties on prediction outputs, using probabilistic methods and statistical inference to account for uncertainties in model parameters and data. The review also discusses the use of data to calibrate and improve turbulence models. Statistical inference is used to account for uncertainties in experimental data and to derive calibrated models that better represent the data. Data-driven modeling approaches, including machine learning, are explored as promising tools for improving turbulence models by learning from large datasets. The review emphasizes the importance of combining data-driven approaches with traditional turbulence modeling techniques to develop more accurate and reliable models. Machine learning is highlighted as a powerful tool for improving turbulence models by learning from data and providing predictions that account for uncertainties. The integration of statistical inference and machine learning offers a promising avenue for developing effective data-driven closures for turbulence models.This review discusses recent developments in turbulence modeling, emphasizing the integration of data-driven approaches with traditional turbulence modeling techniques. Turbulence modeling has traditionally relied on Reynolds-averaged Navier-Stokes (RANS) equations, which require assumptions to close the system of equations. However, these assumptions can introduce inaccuracies, limiting the predictive ability of RANS models. Recent advances in data-driven modeling, statistical inference, and machine learning offer new ways to improve turbulence models by systematically incorporating data to quantify and reduce model uncertainties. The review outlines the four layers of simplifications in RANS closures and discusses the challenges in assessing the predictive nature of turbulence models. It highlights the importance of uncertainty quantification (UQ) in evaluating the reliability of model predictions. UQ aims to rigorously measure and rank the effect of uncertainties on prediction outputs, using probabilistic methods and statistical inference to account for uncertainties in model parameters and data. The review also discusses the use of data to calibrate and improve turbulence models. Statistical inference is used to account for uncertainties in experimental data and to derive calibrated models that better represent the data. Data-driven modeling approaches, including machine learning, are explored as promising tools for improving turbulence models by learning from large datasets. The review emphasizes the importance of combining data-driven approaches with traditional turbulence modeling techniques to develop more accurate and reliable models. Machine learning is highlighted as a powerful tool for improving turbulence models by learning from data and providing predictions that account for uncertainties. The integration of statistical inference and machine learning offers a promising avenue for developing effective data-driven closures for turbulence models.