Aligned, a pioneer in AI infrastructure and scalable machine learning, has announced a strategic collaboration with AMD and the University of Southern California’s Information Sciences Institute (USC ISI) to accelerate the development of MEGALODON, an advanced large language model (LLM). This partnership, revealed on July 24, 2025, harnesses AMD’s Instinct MI300 GPUs and Aligned’s operational expertise to train the model on non-NVIDIA architectures, aiming to redefine AI scalability and efficiency. The initiative is supported by NSF grants and USC’s commitment to open AI research.
Aligned, AMD, and USC ISI collaborate on MEGALODON LLM development.
Uses AMD Instinct MI300 GPUs for non-NVIDIA AI training.
MEGALODON features MEGA architecture for better scalability.
Aligned optimizes training with AMD’s ROCm platform.
Supported by NSF and industry partners like DARPA, Lockheed Martin.
Aims to push AI performance beyond traditional CUDA systems.
MEGALODON, USC ISI’s flagship large language model, introduces a novel Moving Average Equipped Gated Attention (MEGA) architecture, designed to enhance long-context retention while reducing computational complexity. “As researchers push the limits of model size and capability, MEGALODON represents a leap forward in architecture and scalability,” said Jonathan May, research associate professor at the Viterbi School of Engineering. This collaboration leverages AMD’s powerful hardware and Aligned’s infrastructure expertise to train MEGALODON efficiently, marking a significant step in large-scale AI development.
The partnership focuses on optimizing MEGALODON’s training using AMD’s Instinct MI325X GPUs and the ROCm open software platform, a CUDA alternative. “We’re excited to work alongside AMD and USC ISI to help innovate on large-scale AI training,” said Chris Ensey, CEO of Aligned. “Our mission is to enable powerful AI on any hardware—and this partnership exemplifies that vision.” Aligned provides critical tools and engineering support, enabling USC ISI researchers to tap into AMD’s advanced GPU infrastructure, pushing performance boundaries for AI workloads.
MEGALODON’s development is supported by high-performance compute clusters, advanced storage technologies, and access to USC ISI’s MOSIS chip prototyping labs. The project aligns with ISI’s ecosystem of partners, including DARPA, NSF, Lockheed Martin, and Chevron, highlighting its real-world impact. Plans to scale with next-generation AMD GPUs, such as the MI355x, underscore the collaboration’s forward-looking approach to advancing AI research and deployment.
This initiative marks the first major effort to train a large-scale LLM on AMD’s MI325X GPUs via the ROCm ecosystem, emphasizing cross-platform AI enablement. Supported by NSF grants and USC’s commitment to open, interoperable AI research, the collaboration aims to democratize AI development. By integrating the MEGA attention mechanism, MEGALODON offers improved scalability and performance, positioning it as a game-changer in the AI landscape.
The collaboration between Aligned, AMD, and USC ISI represents a bold step toward next-generation AI innovation. By combining cutting-edge hardware, advanced infrastructure, and academic research, the partnership is set to deliver a scalable, efficient, and impactful large language model, paving the way for broader AI adoption across industries.
Aligned specializes in high-performance AI infrastructure, providing advanced tooling, model optimization, and operational expertise across a range of hardware platforms. By supporting frontier models and open research, Aligned is enabling a more robust and equitable AI future.
The Information Sciences Institute (ISI) is a world-renowned research institute within the USC Viterbi School of Engineering whose mission is to make breakthroughs, make an impact. ISI leads groundbreaking research in AI, natural language processing, cybersecurity, and advanced computing. The institute plays a unique role in the R&D ecosystem to turn foundational research breakthroughs into applications and impact.