Learn how to implement agentic RAG in real-world enterprise workflows. Transform you enterprise through:

• Retrieval: How does a RAG agent decide when to search and which sources matter most for your specific question? • Generation: Using large language models, the system generates responses based on the information it has retrieved. • Autonomy: The system makes decisions about when to look for information, what actions to take, and how to handle multi-step tasks.

Modern AI systems are deployed globally, across cultures and in hundreds of languages, yet most safety research and evaluation remains English-centric. In this talk, we will outline a pragmatic roadmap for scaling safety beyond a single linguistic or cultural frame. We will first outline AI safety as a full-stack technical discipline spanning robustness, alignment, privacy, misuse resistance, and critically, evaluation. We will then argue that harm is not universal: what counts as harmful varies with local norms and histories. Drawing on evidence from multilingual red-teaming and jailbreak studies, we will show higher failure rates in low-resource languages and the limits of translate-and-test approaches. We will introduce a global-vs-local harm lens, address data scarcity and long-tail challenges, and present actionable mitigations. Finally, we will examine fairness in model evaluation and close with concrete recommendations for building culturally aware benchmarks and auditing multilingual safety so models are not only capable, but reliably aligned with the communities they serve.

JAX is an ML framework that has been widely adopted by foundation model builders because of its advantages like high performance, scalability, composability, and ease of programmability. In this session, we will showcase the entire ecosystem supported by JAX for end-to-end foundation model building from data loading to training and inference on both TPUs and GPUs. We'll highlight the entire JAX stack, including high performance implementations for large language models and diffusion models in MaxText and MaxDiffusion. Learn how customers such as Assembly AI, Cohere, Anthropic, MidJourney, Stability AI, and partners like Nvidia, have adopted JAX for building foundation models on Google Cloud and beyond.

Click the blue “Learn more” button above to tap into special offers designed to help you implement what you are learning at Google Cloud Next 25.

Implementing generative AI applications with large language models (LLMs), and diffusion models requires large amounts of computation that can seamlessly scale to train, fine-tune, and serve the models. Google Cloud TPUs. Cohere is leveraging the compute-heavy Cloud TPU v4 and v5e to train sophisticated gen AI models that meet the heightened needs of their enterprise users. Check out how Cohere and Cloud TPUs are delivering enterprise-tailored large language models (LLMs) that can help increase business productivity by automating time-consuming and monotonous workflows. Please note: seating is limited and on a first-come, first served basis; standing areas are available

Click the blue “Learn more” button above to tap into special offers designed to help you implement what you are learning at Google Cloud Next 25.

Exploring agents with frameworks like Langchain and vector databases; learn how to build apps that can perform tasks autonomously.

Learn how to tune models on your own treasure trove of data.

Image generators like Midjourney and Stable Diffusion have gotten stronger and stronger in record time. Learn to leverage them for max effect now.

Learn about fine tuning, prompt tuning, guardrails, and middleware to make LLMs more consistent and reliable.

Find out about the latest attack vectors and how LLMs and generative AI present security challenges for teams and how you can mitigate those problems.