Pathway, the data company pioneering live AI that thinks in real-time like humans, has announced that its innovative post-Transformer BDH (Dragon Hatchling) architecture now runs on NVIDIA AI infrastructure and AWS’s cloud and AI technology stack. This integration marks a major advancement in artificial intelligence, enabling a new generation of adaptive and continuously learning AI systems designed for enterprise use. According to Zuzanna Stamirowska, CEO and co-founder of Pathway, this collaboration represents a fundamental shift from static AI models to adaptive intelligence. While traditional Transformer-based large language models have driven significant progress, they remain fixed in their knowledge and behavior. In contrast, BDH’s brain-inspired architecture supports continuous learning, dynamic state management, and Hebbian synaptic plasticity—allowing models to evolve in real time as new data arrives, much like the human brain. The BDH architecture is built for enterprise needs from the ground up, offering greater efficiency, full observability into a model’s live state, and the ability to handle complex, real-time decision-making. It is particularly suited for applications requiring low latency, high adaptability, and deep interpretability—areas where current models fall short. Pathway has chosen AWS as its preferred cloud provider, leveraging its scalable compute infrastructure to deploy BDH at enterprise scale. The NVIDIA Hopper architecture further enhances performance, providing the low-latency, high-throughput environment needed for continuous model adaptation and real-time inference. Jason Bennett, VP and Global Head of Startups and Venture Capital at AWS, highlighted the importance of supporting such innovation. “Enabling innovators to bring new technologies to market is exactly what AWS was created to do,” he said. “We’re excited to partner with Pathway to build AI frameworks that help our customers develop entirely new categories of continuously learning AI applications.” Technically, BDH challenges long-standing assumptions in deep learning. Unlike traditional models that become less interpretable as they grow larger, BDH demonstrates that scale can enhance clarity through neuron specialization. By integrating principles from neuroscience and machine learning, the architecture achieves both high performance and transparency. The model performs competitively on general-purpose hardware and shows strong potential for faster inference on specialized AI processors, offering enterprises lower latency and reduced operational costs. BDH will be available through AWS, with initial design partnerships expected to launch in late 2025. The architecture will be showcased at AWS re:Invent in Las Vegas from December 1 to 5, 2025. Pathway is led by Zuzanna Stamirowska, a complexity scientist, and a team of AI pioneers including CTO Jan Chorowski, who was among the first to apply attention mechanisms to speech and worked with Nobel laureate Geoffrey Hinton at Google Brain. CSO Adrian Kosowski, a quantum physicist and computer scientist, earned his PhD at age 20. The company is backed by prominent investors and advisors, including Lukasz Kaiser, co-author of the Transformer architecture and a key figure in OpenAI’s reasoning models. Headquartered in Palo Alto, California, Pathway is trusted by organizations such as NATO, La Poste, and Formula 1 racing teams. The company is redefining what AI can do—moving beyond static models to systems that learn, adapt, and think in real time.