The LLM-based Generative AI revolution is progressing from the use of language-only models to multimodal models and the transition from monolithic models to more complex Agentic AI workflows. These workflows allow AI systems to address more complex tasks essential to enterprises by doing problem decomposition, planning, self-reflection, and tool use. This talk will share how Intel is collaborating with customers and developers to advance productivity and applications of AI to such higher cognitive tasks using Intel® Gaudi® 3 AI Accelerators, including massive AI cluster buildout in Intel® Tiber™ Developer Cloud.
Vasudev Lal
Principal AI Research Scientist at Intel Labs where I lead the Multimodal Cognitive AI team. The Cognitive AI team develops AI systems that can synthesize concept-level understanding from multiple modalities: vision, language, video, etc. leveraging large-scale AI clusters powered by Intel AI HW (eg: Intel Gaudi-based AI clusters). Vasudev’s current research interests include self-supervised training at scale for continuous and high dimensional modalities like images, video and audio; mechanisms to go beyond statistical learning in today’s AI systems by incorporating counterfactual reasoning and principles from causality and exploring full 3D parallelism (tensor + parallel + data) for training and inferencing large AI models on Intel AI HW (eg: Intel Gaudi-based AI clusters in Intel Dev Cloud). Vasudev obtained his PhD in Electrical and Computer Engineering from the University of Michigan, Ann Arbor in 2012.