9:30 AM–10:00 AM

Check-ins & Breakfast

10:00 AM–10:15 AM

Welcome note 

10:15 AM–11:30 AM

Current state of Generative AI

Generative AI: Hype vs. Reality

Generative AI is one of the most transformative technologies of our time, with the potential to revolutionize many industries. However, it is important to understand the current state of the technology, its limitations, and how to use it responsibly.

This session will provide a comprehensive overview of generative AI, covering the following topics:

• Introducing Gemini: our largest and most capable AI model.
• Current state of generative AI: hype vs. reality: We will explore the latest advances in generative AI, as well as the challenges and limitations that remain.
• Generative AI terminology: We will define the key terms used in generative AI, such as foundation model, prompt, and vector databases.
• Limitations of generative AI: We will discuss the main limitations of generative AI, such as bias, accuracy, and safety.
• Overview of AI on Google Cloud, including 1P and 3P foundation models: We will provide an overview of the AI services offered by Google Cloud, including foundation models from Google and third-party providers.
• When to use and not use LLMs: We will discuss the best practices for using large language models (LLMs), as well as the cases where LLMs should not be used.
• Responsible AI and ethics: We will explore the ethical and social implications of generative AI, and how to use it responsibly.

11:30 AM–12:00 PM

Break & Refreshments 

12:00 PM–1:20 PM

Exploring prompt engineering

Generative AI foundation models are powerful tools that can be used to generate text, code, and embeddings. How can you get started? This session will provide a comprehensive introduction to using the text, chat, code, and embedding foundation models using the Python SDK.

We will cover the following topics:

• Introduction to generative AI foundation models: We will discuss the different types of generative AI foundation models, their capabilities
• Using the Python SDK to call the foundation models: We will show you how to use the Python SDK to call the text, chat, code, and embedding foundation models, along with various parameters.

Best practices for prompt engineering: We will discuss the best practices for writing prompts that will generate the desired outputs from the foundation models. 

1:20 PM–1:30 PM

Closing Remarks

9:30 AM–10:00 AM

Check-ins & Breakfast

10:00 AM–10:15 AM

Welcome note

10:15 AM–11:30 AM

Deep dive into embeddings and vector databases

Embeddings are a powerful tool for representing data in a way that can be used by machine learning models. In this seminar, we will cover the following topics:

• What are embeddings? We will discuss the different types of embeddings and how they are created.
• Why are embeddings useful? We will explore the many benefits of using embeddings, such as their ability to reduce dimensionality, capture semantic relationships, and improve the performance of machine learning models.
• Using embeddings with vector databases: We will show you how to use vector databases to store and query embeddings.
• Applications of embeddings: We will discuss some of the most common applications of embeddings, such as similarity search and retrieval augmented generation.

11:30 AM–12:00 PM

Break & Refreshments 

12:00 PM–1:20 PM

Intro to RAG architectures including Vertex AI Search 

What is the latest buzz around Retrieval Augmented Generation (RAG)? RAG enables the  access and retrieval of information from external knowledge bases, which allows them to generate more accurate, informative, and up-to-date responses.

This session will provide a comprehensive overview of RAG, covering the following topics:

• What is retrieval augmented generation? We will discuss the basic concepts and principles of RAG.
• RAG architectures: Best practices towards designing RAG architectures for storage, retrieval and ranking
• Managed solutions (e.g. Vertex AI Search and Conversation): We will discuss how to use managed solutions like Vertex AI Search and Conversation that make it easy to build and deploy RAG models.

1:20 PM–1:30 PM

Closing Remarks

9:30 AM–10:00 AM

Check-ins & Breakfast

10:00 AM–10:15 AM

Welcome note

10:15 AM–11:30 AM

Building enterprise chat apps using GenAI

In the context of communication, convergent thinking focuses on reaching one well-defined solution to a problem as opposed to divergent thinking which involves more creativity.  

When designing virtual agents, we want to ensure that we are divergent when it comes to options and possibilities, and convergent when we are trying to help users solve a problem or make transactions. 

In this class and the associated Qwiklabs you will explore the different tools and services for working with Conversational AI in Google Cloud. 

Specifically you will learn how to:

• Design a blend of deterministic goal-oriented conversation and how to switch to a generative and LLM-based approach when it’s appropriate.
• Use features that help ground your generative responses in your own data, or features that help make free-form calls to the PaLM API at certain points in the conversation.
• Design your agent to cover the most common paths (followed by 80% users) and let LLMs handle the remaining 20% of paths that include edge cases and common detours.
• Use new ways for creating Dialogflow CX agents using LLMs where you provide natural language instructions and structured data in the form of playbooks.

11:30 AM–12:00 PM

Break & Refreshments

12:00 AM–1:20 PM

Deploying and hosting apps in the cloud

Generative AI Seminar: Deploying Generative AI Apps on Google Cloud

Generative AI is rapidly transforming many industries, and Google Cloud offers a wide range of tools and services for deploying Generative AI applications. This session will cover the following topics:

• Deploying Generative AI apps on Google Cloud: Learn how to deploy Generative AI applications to Google Cloud using popular frameworks like Flask, Gradio, and Streamlit.
• Building and deploying chat apps to Cloud Run: Learn how to build and deploy chat apps to Cloud Run, a serverless compute platform that makes it easy to build and run stateless containers.
• Deploying Cloud Functions that trigger Generative AI workflows: Learn how to deploy Cloud Functions, which are event-driven compute resources that can be used to trigger Generative AI workflows based on events in Cloud Storage, such as when a new file is uploaded.

1:20 PM–1:30 PM

Closing Remarks

9:30 AM–10:00 AM

Check-ins & Breakfast

10:00 AM–10:15 AM

Welcome note

10:15 AM–11:30 AM

Tuning & RLHF

Foundation models are powerful language models that can be used to perform a wide range of tasks, including translation, summarization, and code generation. However, these models are often trained on general-purpose data and may not perform well on specific tasks. Tuning a foundation model can help to improve its performance on a specific task by fine-tuning its parameters on a smaller dataset of labeled data.

This session  will cover the following topics:

• What is foundation model tuning?
• How to tune foundation models
• When and when not to use tuning
• What is reinforcement learning with human feedback (RLHF)?
• How to use RLHF to tune foundation models

11:30 AM–12:00 PM

Break & Refreshments

12:00 AM–1:20 PM

Deploy Open Source on Vertex AI with Mistral AI

Open Source models are a great alternative for flexibility and added customization. 

Discover Mistral AI open source models, learn how to deploy Mistral 7B and Mixtral 8x7B on Vertex AI to build your own custom architectures, RAG and fine-tuning! 

1:20 PM–1:30 PM

Q/A & Closing Remarks

1:30 PM–3:00 PM

Graduation Ceremony