Mesh-free particle methods such as Smoothed Particle Hydrodynamics (SPH) and Moving Particle Semi-Implicit (MPS) have gained significant attention for continuum modeling due to their mesh-free, Lagrangian nature, which allows for the flexible handling of highly deformed boundaries and interfaces. However, the uncontrolled particle motion can affect stability, accuracy, and convergence in these methods. Additionally, computational performance, boundary condition implementation, and incomplete or uncertain physics (such as rheology and turbulence) remain key challenges in certain applications. Recent advances in Scientific Machine Learning (Sci-ML) have demonstrated great potential in enhancing and accelerating physics-based numerical methods. In this work, we explore how machine learning (ML) techniques can address some of the challenges in particle methods. Specifically, we present our recent works in developing and evaluating ML-based techniques for enhancing accuracy and stability, improving boundary condition implementation, and modeling complex rheological behavior using particle methods.