Keysight Technologies has announced the launch of its KAI Data Center Builder, a software suite designed to emulate real-world workloads for artificial intelligence (AI) applications. This solution aims to evaluate how new algorithms, components, and protocols can enhance the performance of AI training systems. By reproducing network communication patterns and integrating large language model (LLM), the KAI Data Center Builder allows users to optimize AI workloads and fine-tune infrastructure without the need for expensive, large-scale deployments.

The software enables AI operators to experiment with various parallel processing strategies, adjust model partitioning, and analyze factors such as network utilization and latency. As a result, users can gain insights into bottlenecks and improve training efficiency. Keysight emphasizes the need for full-stack validation and optimization in AI infrastructure to mitigate costly delays during design and manufacturing phases.

Ram Periakaruppan, Vice President and General Manager of Network Test & Security Solutions at Keysight, stated that KAI Data Center Builder's workload emulation capability delivers a realistic environment for validating and optimizing AI component designs. This development aligns with the KAI architecture, which is focused on helping customers scale their AI processing in data centers.

Keysight plans to showcase the KAI Data Center Builder at the upcoming OFC 2025 conference in San Francisco.