Mark Oskin

Papers

Total Citations

H-Index

About

Mark Oskin is a pioneering computer architect whose research centers on memory systems, near-data processing, and energy-efficient computing. His most influential work, "NCAM: Near-Data Processing for Nearest Neighbor Search" (2016), tackles a critical bottleneck in modern AI workloads: the data movement that slows k-nearest neighbor (kNN) search, a core algorithm in natural language processing, computer vision, and robotics. By proposing a near-data processing architecture that moves computation closer to memory, Oskin demonstrated how to dramatically reduce latency and improve throughput for these memory-bound applications. This contribution has earned 4 citations and represents a key step in rethinking the memory hierarchy for machine learning. Beyond this paper, Oskin is known for his broader work on processor design, including the influential "Carbon" architecture for low-power computing and contributions to the development of the RISC-V instruction set. His research has shaped how modern systems handle data-intensive tasks, making him a respected voice in the architecture community. For students and researchers, Oskin’s work offers a compelling example of how reexamining fundamental system bottlenecks can unlock new performance frontiers.