https://blog.chameleoncloud.org/authors/xiaoyang-lu/Recent content in Xiaoyang Lu on Blog | ChameleonHugoen-usMon, 29 Jun 2026 00:00:00 +0000https://blog.chameleoncloud.org/posts/attenio-io-analysis-long-sequence-attention/Mon, 29 Jun 2026 00:00:00 +0000https://blog.chameleoncloud.org/posts/attenio-io-analysis-long-sequence-attention/<p>Modern AI systems are often limited not only by how much computation they perform, but also by how much data they need to move. Moving data — or I/O — between large off-chip memory and small on-chip memory can be expensive, and this cost becomes especially important for data-intensive AI workloads. Xiaoyang Lu, a Research Assistant Professor at Illinois Institute of Technology, has been investigating how to attack this problem at its root, applying I/O analysis to design hardware that minimizes data movement rather than simply maximizing compute throughput.</p>