Ori Goshen, AI21 Labs
Artificial intelligence factories focus on building specialized cost-efficient AI systems for business by fine-tuning foundation models and integrating proprietary data into production workflows. Ori Goshen of AI21 Labs is co-founder and chief executive officer. Goshen explains how AI21 shifts from building generic foundation models to delivering lab-as-a-service and platform tools that fine-tune open models for specific business workflows. They outline approaches to model training, inference optimization and deploying production-ready specialized AI systems. This theCUBE Research segment is hosted by Gemma Allen of theCUBE and NYSE Wired. Key takeaways include the economic case for specialization and ownership. Goshen emphasizes that fine-tuning open models and embedding proprietary data into model weights creates a competitive flywheel and continuous improvement. They highlight precision compute orchestration to reduce cost and cite Deel's payroll model as achieving frontier-level performance at roughly 50x lower cost. They also note AI21's partnership with NVIDIA as a strategic enabler for post-training deployments and production scale. The discussion covers model fine-tuning, foundation models, compute optimization, inference optimization and AI infrastructure considerations for data centers.
Share this session