The video features a conversation at Advancing AI 2025 in San Jose between a host from AMD and Steen Graham, the founder of Metrum AI. Steen shares the journey of Metrum AI, which he started about four years ago, initially focusing on deep learning and convolutional neural networks before the rise of transformer models. The advent of transformer models significantly accelerated their work, enabling them to develop advanced AI agents. Steen emphasizes the strong partnership with AMD, appreciating AMD’s open and programmable technology stack that fosters innovation and flexibility, unlike more vertically integrated competitors.
Steen highlights AMD’s hardware capabilities, particularly the best-in-class CPUs and the MI300X accelerators, which provide the necessary memory footprint and cost-efficiency for running complex AI workloads. Metrum AI leverages AMD’s EPYC processors and Instinct accelerators to run multiple AI models simultaneously, including vision-language models, large language models, audio models, and segmentation models. This multi-model approach is essential for their AI agents, which are rebuilt frequently to keep pace with rapid innovation, focusing on delivering functional outcomes rather than just raw speed.
The core applications of Metrum AI’s technology revolve around infrastructure monitoring and automation. Steen explains how their AI agents autonomously monitor data centers, manage incidents, and maintain regulatory compliance, reducing downtime and improving customer satisfaction. They also deploy AI agents in cybersecurity to provide continuous defense against evolving threats by automating detection and remediation processes. Beyond data centers, Metrum AI’s agents support manufacturing and telecommunications operations, analyzing quality of service metrics and optimizing infrastructure performance.
Looking ahead, Steen discusses Metrum AI’s vision to empower enterprises with customizable, secure AI agents that require little to no coding. They are developing products that enable organizations to configure AI agents tailored to their proprietary data, workflows, and APIs, simplifying AI adoption and integration. This no-code approach aims to democratize AI technology, allowing businesses to harness AI’s power without needing extensive technical expertise, thereby accelerating digital transformation across industries.
The conversation concludes with mutual appreciation, with the AMD host praising Steen’s passion and vision for AI innovation. Steen expresses gratitude for AMD’s hardware and open software ecosystem, which he credits as foundational to Metrum AI’s success. The partnership exemplifies how cutting-edge hardware and flexible software platforms can drive the next wave of AI agents, delivering practical solutions that enhance operational efficiency and security across various sectors.