The objective of the proposed research is to investigate the Memory system optimizations for parallel and bandwidth-intensive applications. Memory system performance is a key factor in determining the overall system performance and cost of modern datacenters and cloud systems. Driven by the applications’ demand for higher throughput and performance there has been a surge for higher memory capacity and bandwidth while keeping the energy and monetary costs low. Further, with the slowdown of Moore’s law, packaging limitations, and signaling challenges posed by conventional memory systems, scaling the memory capacity and bandwidth has been a challenge to address the needs of contemporary workloads. Although, the simultaneous enhancement of all three aspects – Capacity, Bandwidth, and Cost – presents a formidable challenge, however, there exists a promising avenue for augmenting application performance with the implementation of judicious memory system optimizations tailored to the specific requirements of workloads and the underlying hardware systems deployed for computation. In my proposal, I plan to demonstrate the performance enhancements achieved through memory system optimizations for accelerating parallel and scientific workloads on resource-constrained Hardware Transactional Memory Systems, and throughput improvements obtained by expanding per-node memory capacity and bandwidth for Deep Learning training applications. Additionally, I will present my ongoing work on boosting the performance for bandwidth-intensive applications on high-performance servers utilizing CXL-based memory expansion techniques.