Accelerate your generative AI projects with new tools on Vertex AI
TLDRGoogle Cloud's Jason Gelman and Estée Lauder's Eric Higgins discuss Vertex AI's generative capabilities for accelerating AI projects. They cover the Model Garden for AI model starting points, Estée Lauder's use of Google Cloud for luxury beauty solutions, and Vertex AI's role in enhancing customer experiences through personalized recommendations and generative AI. New tools for model tuning, distillation, and evaluation are highlighted, emphasizing the importance of responsible AI practices.
Takeaways
- 🚀 Google Cloud's Vertex AI has been enhanced with new tools to accelerate generative AI projects, providing a suite of products including AI capabilities for business solutions.
- 🌟 The Model Garden is a curated collection of models, offering a starting point for AI journeys, with new models and extensions added this year.
- 📈 Estée Lauder has leveraged Google Cloud and Vertex AI to deliver AI-enabled solutions, enhancing customer experiences and operational efficiency in the luxury beauty space.
- 🛠️ The use of open-source tools and a DevOps culture has been crucial for Estée Lauder's data science team, enabling flexibility and scalability in their AI projects.
- 📊 Early successes for Estée Lauder included A/B testing frameworks, advanced recommendation systems, and customer segmentation models, leading to cost savings and spend efficiency.
- 🤖 Vertex AI's integration into the workflow allowed Estée Lauder to focus on data science solutions and business value, with reduced time spent on infrastructure concerns.
- 🌐 The architecture of Estée Lauder's data science platform is built on BigQuery and includes various Google Cloud tools for CI/CD pipelines and machine learning.
- 💡 Vertex AI's Model Garden now includes 40 foundation models, 60 open-source models, and is continuously growing, offering a wide range of models for different modalities and tasks.
- 🔍 Model Garden provides end-to-end journeys, allowing users to find, tune, and deploy models with ease, and integrates with MLOps tooling for generative AI and other models.
- 🌟 New features in Vertex AI include streaming responses, enabling real-time generation and reduced latency, and Grounding, which customizes model responses to enterprise data.
Q & A
What is the main focus of the session on Vertex AI and its generative capabilities?
-The main focus of the session is to discuss how Vertex AI and its generative capabilities can be used to accelerate AI projects and deliver AI-enabled solutions for business stakeholders.
Who is Jason Gelman and what is his role at Google Cloud?
-Jason Gelman is the individual who manages several AI projects at Google Cloud. He provides an overview of Vertex AI and its place within the suite of Google Cloud products, including the newly released generative AI capabilities.
What is the significance of the Model Garden in Vertex AI?
-The Model Garden is a curated set of models, including open-source, Google-first party, and third-party models, that serves as a starting point for AI journeys. It is an integral part of Vertex AI, providing a foundation for training, hosting, tuning, distilling, and evaluating models.
How does Estée Lauder utilize Vertex AI to enhance its business operations?
-Estée Lauder uses Vertex AI to implement AI-enabled solutions that cater to its diverse brands and regions. These solutions include A/B testing frameworks, advanced recommendation systems, consumer lifetime value models, and customer segmentation, all aimed at improving operational efficiency and driving commercial excellence.
What are some of the key challenges faced by Estée Lauder in the luxury beauty space?
-Estée Lauder faces challenges such as fierce competition, the need for personalized and omnichannel consumer experiences, keeping up with ever-changing trends in the beauty and fashion industries, and maintaining the unique brand identity and voice of each of its brands.
How has the partnership with Eviden helped Estée Lauder in its data science initiatives?
-Eviden, as a development partner, has provided staff augmentation to help Estée Lauder scale its resources. They have also contributed essential knowledge of Google Cloud products, furthering the company's data engineering, analytics, and machine learning efforts.
What is the role of BigQuery in Estée Lauder's analytics initiatives?
-BigQuery plays a significant role in Estée Lauder's analytics initiatives by forming the backbone of their BI and total data infrastructure. It enables the company to manage and analyze large volumes of data, contributing to informed decision-making and strategic planning.
What are some of the generative AI use cases explored by Estée Lauder?
-Estée Lauder has explored generative AI use cases such as improving customer call log management through LLMs for better labeling and issue identification, producing reviews digests to quickly summarize consumer feedback, and accelerating creative workflows through copy-related activities like variant generation.
How does Vertex AI support the development of responsible AI?
-Vertex AI integrates responsible AI tooling, including content moderation APIs, bias evaluation tools, and other safeguards to ensure that AI models are developed and deployed ethically and in compliance with global regulations. It also provides enterprise controls to maintain data security and privacy.
What is the significance of the new OSS models integrated into Model Garden?
-The integration of new OSS models into Model Garden, such as those from Meta, expands the range of models available to users. This provides more choices for the best models suited to specific workloads and ensures that users have access to the latest open-source model releases.
How does the Model Garden facilitate the deployment of AI models?
-Model Garden simplifies the deployment process by allowing users to find, tune, and deploy models to custom endpoints within a few clicks. It integrates various models with Vertex's powerful tooling, offering a seamless experience for users to develop and deploy AI solutions.
Outlines
📣 Introduction to Vertex AI and Generative Capabilities
The session begins with Jason Gelman welcoming everyone to the presentation on Vertex AI and its generative capabilities. He introduces himself as the AI project manager at Google Cloud and is joined by Eric, the vice president of data science at Estée Lauder. The agenda includes an overview of Vertex AI, its integration into Google Cloud's suite of products, and how Estée Lauder utilizes Vertex AI for business solutions. Eric discusses the challenges faced by Estée Lauder in the luxury beauty space, emphasizing the importance of clienteling, online experiences, and keeping up with trends. He also highlights the company's guiding principles and the establishment of a data science organization.
🌟 Estée Lauder's Data Science Journey with Vertex AI
Eric Higgins shares the story of Estée Lauder's data science team formation in 2021, coinciding with the launch of Vertex AI. The team's mission is to deliver results-oriented data science solutions across all brands. Eric emphasizes the need for flexibility due to the diverse operations of the brands globally. He mentions the importance of DevOps discipline in their machine learning development, which has been significantly enabled by Vertex AI and Google Cloud tools. The team initially used open-source platforms like Kubeflow pipelines and progressively integrated more Google Cloud services, leading to early wins in A/B testing frameworks, recommendation systems, and consumer lifetime value models.
🛠️ Leveraging Vertex AI for Business Value and Efficiency
The discussion continues with Eric explaining how Vertex AI has been central to their operations, allowing the team to focus on business deliverables. He mentions the collaboration with Eviden, a development partner, which helped scale resources and provided essential knowledge in data engineering and machine learning. Eric presents an architecture diagram showing Vertex AI's role and the integration of BigQuery for analytics initiatives. He also talks about the use of Google Artifact Registry for CI/CD tooling. Estée Lauder has been deploying new recommendation tooling from Vertex AI to enhance personalization on their websites. The team has also explored generative AI for customer call log analysis and review summarization, improving issue identification and consumer experience.
🌱 Model Garden: A Rich Collection of AI Models
Jason Gelman takes over to discuss the Model Garden, Google Cloud's collection of the best Google, open-source, and partner models integrated with Vertex AI's tooling. He explains that Model Garden enables users to find, tune, and deploy models for their tasks easily. The collection includes a variety of large foundation models and more task-specific ones, with 40 foundation models and 60 open-source models. Jason emphasizes the commitment to providing customers with choice and the seamless integration of these models into Vertex. He also mentions the addition of Meta's models, like Llama 2 and Code Llama, and the pre-announcement of Claude 2 from Anthropic.
🚀 Demonstrating Model Garden's Capabilities
Jason proceeds to demonstrate Model Garden's capabilities with a live demo, showcasing how it can be used to generate content for a healthy snack product line. He explains the process of using PaLM 2 for text generation, adjusting settings like temperature for creativity, and token limit. He also introduces new features like streaming and grounding, which allow for real-time output and customization against enterprise data. Despite a technical issue with the internet connection, Jason manages to convey the ease of using Model Garden for content generation, tuning, and deployment.
🔍 Enhancing Product Search with AI
In the final paragraph, Jason discusses the use of AI for enhancing product search on e-commerce websites. He introduces the concept of text embeddings for representing product descriptions in a vector space to understand similarity. He explains the two-step process involving Model Garden's text embedding model and the Vertex AI Matching Engine, a managed vector database. Jason suggests using this combination for robust product search, mentioning the recent release of an image embeddings model for searching based on images. He concludes by encouraging the audience to explore Model Garden and assures that their accounts will work effectively.
Mindmap
Keywords
💡Vertex AI
💡Generative AI
💡Estée Lauder
💡Model Garden
💡AI Projects
💡Data Science
💡Machine Learning
💡DevOps
💡Responsible AI
💡Cloud Computing
Highlights
Google Cloud's Vertex AI offers a suite of tools to accelerate generative AI projects.
Vertex AI integrates with newly released generative AI capabilities within Google Cloud's product suite.
Estée Lauder collaborates with Vertex AI to deliver AI-enabled solutions for business stakeholders.
Google Cloud provides a complete set of capabilities and solutions, including infrastructure and AI platform extensions.
Model Garden serves as a starting point for AI journeys, offering a curated set of open-source and Google-first-party models.
Vertex AI's capabilities have been extended to include tuning, distilling, and evaluating models.
Estée Lauder has utilized Vertex AI for A/B testing frameworks, reducing costs and increasing test speed.
Advanced recommendation systems and consumer lifetime value models have been implemented by Estée Lauder to increase marketing efficiency.
Estée Lauder's data science team was formed in 2021, coinciding with the launch of Vertex AI, to deliver results-oriented solutions.
The use of open-source tools and GCP services has been integral to Estée Lauder's data science platform development.
Eviden, a development partner, has helped Estée Lauder scale resources and further data engineering and machine learning efforts.
Vertex AI plays a central role in Estée Lauder's architecture, simplifying processes and focusing on business deliverables.
Estée Lauder uses Vertex AI to enhance personalization and recommendations on their brand websites.
Generative AI has been explored by Estée Lauder for uses such as customer call log analysis and review summarization.
LLMs (Large Language Models) have been utilized to accelerate creative workflows and copy-related activities within Estée Lauder.
Model Garden integrates 40 foundation models, 60 open-source models, and various task-specific models.
Model Garden provides enterprise-level controls and responsible AI tooling integrated with all models.
Google is committed to providing choice and integration of the best models for customer workloads in Model Garden.
New OSS (Open Source Software) models and partner models are continuously integrated into Model Garden.
