Unlock the power of Oracle Database 23AI and Autonomous Database Serverless (ADB-S) with this comprehensive guide to the latest innovations in performance, security, automation, and AI-driven optimization. As enterprises embrace intelligent and autonomous data platforms, understanding these capabilities is essential for data architects, developers, and DBAs. Explore cutting-edge features such as vector data types and AI-powered vector search, revolutionizing data retrieval in modern AI applications. Learn how schema privileges and the DB_DEVELOPER_ROLE simplify access control in multi-tenant environments. Dive into advanced auditing, SQL Firewall, and data integrity constraints to strengthen security and compliance. Discover AI-driven advancements like machine learning-based query execution, customer retention prediction, and AI-powered query tuning. Additional chapters cover innovations in JSON, XML, JSON-Relational Duality Views, new indexing techniques, SQL property graphs, materialized views, partitioning, lock-free transactions, JavaScript stored procedures, blockchain tables, and automated bigfile tablespace shrinking. What sets this book apart is its practical focus—each chapter includes real-world case studies and executable scripts, enabling professionals to implement these features effectively in enterprise environments. Whether you're optimizing performance or aligning IT with business goals, this guide is your key to building scalable, secure, and AI-powered solutions with Oracle 23AI and ADB-S. What You Will Learn Explore Oracle 23AI's latest features through real-world use cases Implement AI/ML-driven optimizations for smarter, autonomous database performance Gain hands-on experience with executable scripts and practical coding examples Strengthen security and compliance using advanced auditing, SQL Firewall, and blockchain tables Master high-performance techniques for query tuning, in-memory processing, and scalability Revolutionize data access with AI-powered vector search in modern AI workloads Simplify user access in multi-tenant environments using schema privileges and DB_DEVELOPER_ROLE Model and query complex data using JSON-Relational Duality Views and SQL property graphs Who this Book is For Database architects, data engineers, Oracle developers, and IT professionals seeking to leverage Oracle 23AI’s latest features for real-world applications

Discover how leading enterprises are leveraging 15+ years of Amazon RDS operational excellence to power their SQL Server and Oracle databases in the cloud. In this session, explore features across Amazon RDS for SQL Server and Oracle that help you achieve substantial cost savings, enhanced scalability, and efficient operations. Through real-world cost optimization techniques and architectural best practices, learn how organizations are reducing operational overhead and costs while improving availability, scalability, and performance.

Learn more: More AWS events: https://go.aws/3kss9CP

Subscribe: More AWS videos: http://bit.ly/2O3zS75 More AWS events videos: http://bit.ly/316g9t4

ABOUT AWS: Amazon Web Services (AWS) hosts events, both online and in-person, bringing the cloud computing community together to connect, collaborate, and learn from AWS experts. AWS is the world's most comprehensive and broadly adopted cloud platform, offering over 200 fully featured services from data centers globally. Millions of customers—including the fastest-growing startups, largest enterprises, and leading government agencies—are using AWS to lower costs, become more agile, and innovate faster.

AWSreInvent #AWSreInvent2025 #AWS

CLIP is a foundational model with transferable classification performance in the few-shot setting. Several methods have shown improved performance of CLIP using few-shot examples. However, so far, all these techniques have been benchmarked using standard few-shot datasets. We argue that this mode of evaluation does not provide a true indication of the inductive generalization ability using few-shot examples. As most datasets have been seen by the CLIP model, the resultant setting can be termed as partially transductive. To solve this, we propose a pipeline that uses an unlearning technique to obtain true inductive baselines. In this new inductive setting, the methods show a significant drop in performance (-55% on average among 13 baselines with multiple datasets). We validate the unlearning technique using oracle baselines. An improved few-shot classification technique is proposed that consistently obtains state-of-the-art performance over 13 other recent baseline methods on a comprehensive analysis with 5880 experiments - varying the datasets, differing number of few-shot examples, unlearning setting, and with different seeds. Thus, we identify the issue with the evaluation of CLIP-based few-shot classification, provide a solution using unlearning, propose new benchmarks, and provide an improved method.

CLIP is a foundational model with transferable classification performance in the few-shot setting. Several methods have shown improved performance of CLIP using few-shot examples. However, so far, all these techniques have been benchmarked using standard few-shot datasets. We argue that this mode of evaluation does not provide a true indication of the inductive generalization ability using few-shot examples. As most datasets have been seen by the CLIP model, the resultant setting can be termed as partially transductive. To solve this, we propose a pipeline that uses an unlearning technique to obtain true inductive baselines. In this new inductive setting, the methods show a significant drop in performance (-55% on average among 13 baselines with multiple datasets). We validate the unlearning technique using oracle baselines. An improved few-shot classification technique is proposed that consistently obtains state-of-the-art performance over 13 other recent baseline methods on a comprehensive analysis with 5880 experiments - varying the datasets, differing number of few-shot examples, unlearning setting, and with different seeds. Thus, we identify the issue with the evaluation of CLIP-based few-shot classification, provide a solution using unlearning, propose new benchmarks, and provide an improved method.

CLIP is a foundational model with transferable classification performance in the few-shot setting. Several methods have shown improved performance of CLIP using few-shot examples. However, so far, all these techniques have been benchmarked using standard few-shot datasets. We argue that this mode of evaluation does not provide a true indication of the inductive generalization ability using few-shot examples. As most datasets have been seen by the CLIP model, the resultant setting can be termed as partially transductive. To solve this, we propose a pipeline that uses an unlearning technique to obtain true inductive baselines. In this new inductive setting, the methods show a significant drop in performance (-55% on average among 13 baselines with multiple datasets). We validate the unlearning technique using oracle baselines. An improved few-shot classification technique is proposed that consistently obtains state-of-the-art performance over 13 other recent baseline methods on a comprehensive analysis with 5880 experiments - varying the datasets, differing number of few-shot examples, unlearning setting, and with different seeds. Thus, we identify the issue with the evaluation of CLIP-based few-shot classification, provide a solution using unlearning, propose new benchmarks, and provide an improved method.

CLIP is a foundational model with transferable classification performance in the few-shot setting. Several methods have shown improved performance of CLIP using few-shot examples. However, so far, all these techniques have been benchmarked using standard few-shot datasets. We argue that this mode of evaluation does not provide a true indication of the inductive generalization ability using few-shot examples. As most datasets have been seen by the CLIP model, the resultant setting can be termed as partially transductive. To solve this, we propose a pipeline that uses an unlearning technique to obtain true inductive baselines. In this new inductive setting, the methods show a significant drop in performance (-55% on average among 13 baselines with multiple datasets). We validate the unlearning technique using oracle baselines. An improved few-shot classification technique is proposed that consistently obtains state-of-the-art performance over 13 other recent baseline methods on a comprehensive analysis with 5880 experiments - varying the datasets, differing number of few-shot examples, unlearning setting, and with different seeds. Thus, we identify the issue with the evaluation of CLIP-based few-shot classification, provide a solution using unlearning, propose new benchmarks, and provide an improved method.

CLIP is a foundational model with transferable classification performance in the few-shot setting. Several methods have shown improved performance of CLIP using few-shot examples. However, so far, all these techniques have been benchmarked using standard few-shot datasets. We argue that this mode of evaluation does not provide a true indication of the inductive generalization ability using few-shot examples. As most datasets have been seen by the CLIP model, the resultant setting can be termed as partially transductive. To solve this, we propose a pipeline that uses an unlearning technique to obtain true inductive baselines. In this new inductive setting, the methods show a significant drop in performance (-55% on average among 13 baselines with multiple datasets). We validate the unlearning technique using oracle baselines. An improved few-shot classification technique is proposed that consistently obtains state-of-the-art performance over 13 other recent baseline methods on a comprehensive analysis with 5880 experiments - varying the datasets, differing number of few-shot examples, unlearning setting, and with different seeds. Thus, we identify the issue with the evaluation of CLIP-based few-shot classification, provide a solution using unlearning, propose new benchmarks, and provide an improved method.

CLIP is a foundational model with transferable classification performance in the few-shot setting. Several methods have shown improved performance of CLIP using few-shot examples. However, so far, all these techniques have been benchmarked using standard few-shot datasets. We argue that this mode of evaluation does not provide a true indication of the inductive generalization ability using few-shot examples. As most datasets have been seen by the CLIP model, the resultant setting can be termed as partially transductive. To solve this, we propose a pipeline that uses an unlearning technique to obtain true inductive baselines. In this new inductive setting, the methods show a significant drop in performance (-55% on average among 13 baselines with multiple datasets). We validate the unlearning technique using oracle baselines. An improved few-shot classification technique is proposed that consistently obtains state-of-the-art performance over 13 other recent baseline methods on a comprehensive analysis with 5880 experiments - varying the datasets, differing number of few-shot examples, unlearning setting, and with different seeds. Thus, we identify the issue with the evaluation of CLIP-based few-shot classification, provide a solution using unlearning, propose new benchmarks, and provide an improved method.

For years, businesses have relied on Microsoft to power critical systems and applications like VMware, SAP, and Oracle. With Azure, organizations can now modernize these systems more efficiently and securely, enabling innovation at their own pace. Explore partner strategies for transforming workloads and see how leveraging Microsoft’s tools and investments can accelerate success and uncover new opportunities.

Learn how to embed AI across your Oracle estate, spanning apps and databases with Oracle Database@Azure. We’ll walk through blueprints that use Microsoft Foundry, Copilot Studio, and Microsoft Fabric to vectorize Oracle data, build task?specific copilots and agents, and operationalize analytics. Hear customer?proven integration patterns, cost considerations, and reliability tips, and leave with a pragmatic adoption roadmap and next?step resources.

Apollo Hospitals unlocked faster performance and sped up innovation by migrating to Azure Database for PostgreSQL. Ready to move beyond your legacy databases? Join us to learn how to make the move seamless with AI-assisted migration for Oracle workloads. Plus, hear how a move to Azure enables innovation and explore the latest features that deliver the resilience and extensibility needed to bolster your databases to support more intelligent applications and demanding workloads.

Equinor, a $100B global energy leader partnered with Tessell to migrate 700+ Oracle and SQL Server databases to Azure - without re-architecting. With cloud-native HA/DR, automation, and cost visibility, they achieved 65% efficiency gains and cut provisioning from weeks to hours. Not theory: this is a real, enterprise-tested approach to modernizing mission-critical databases at scale. It’s secure, efficient, and built for the cloud.

Join us to hear how Levi’s transformed its global IT estate—including Windows Server, SQL Server, SAP, Oracle, and more—through a comprehensive migration and modernization strategy. Learn how they combined data and app modernization with infrastructure re-platforming to unlock agility, scalability, and cost savings. This real-world story will highlight the strategy, lessons learned, and tangible business outcomes that came from taking an end-to-end approach to migration and modernization.

Are your Oracle applications and databases stuck on-premises? Azure makes it easy to unlock their full potential by bringing Oracle workloads to the cloud, running them alongside your other cloud applications on the same high-performance, secure network. Unify Oracle data with other critical sources using Microsoft Fabric, and accelerate app modernization with the power of Azure AI services. Discover how to streamline operations, boost agility, and maximize the value of your Oracle investments.

Oracle Database@Azure provides the performance and flexibility of using Autonomous AI Database and Exadata Database Service, enabling you to build and deploy powerful Oracle AI Database apps natively in Microsoft Azure. In this session, we’ll discuss how the converged capabilities of Oracle AI Database, including AI Vector Search, combine with Azure AI Services to help accelerate the development of advanced AI-driven applications.

This new technology area encompasses a wide array of mission-critical workloads and modernization efforts, including SAP, Oracle Database@Azure, Azure VMware Solution (AVS), Nutanix on Azure, and mainframe modernization strategies. MVPs in this area are expected to demonstrate deep expertise in at least one of these technologies, with a strong grasp of architecture, deployment, and operational nuances on Azure.

Connection Pods accommodate up to 15 people. Please RSVP and arrive at least 5 minutes before the start time, at which point remaining spaces are open to standby attendees.

You probably don’t need a collection of specialty databases. Just use Postgres instead! Written for application developers and database pros, Just Use Postgres! shows you how to get the most out of the powerful Postgres database. In Just Use Postgres! you’ll learn how to: Use Postgres as an RDBMS for transactional workloads Develop generative AI, geospatial, and time-series applications Take advantage of modern SQL including window functions and CTEs Perform full-text search and process JSON documents Use Postgres as a message queue Optimize performance with various index types including B-trees, GIN, GiST, HNSW, and more Over the decades, PostgreSQL, aka Postgres, has grown into the most powerful general-purpose database and has become the de facto standard for developers worldwide. Just Use Postgres! takes a modern look at Postgres, exploring the database’s most up-to-date features for AI, time-series, full-text search, geospatial, and other application workloads. About the Technology You know that PostgreSQL is a fast, reliable, SQL compliant RDBMS. You may not know that it’s also great for geospatial systems, time series, full-text search, JSON documents, AI vector embeddings, and many other specialty database functions. For almost any data task you can imagine, you can use Postgres. About the Book Just Use Postgres! covers recipes for using Postgres in dozens of applications normally reserved for single-purpose databases. Written for busy application developers, each chapter explores a different use case illuminating the breadth and depth of Postgres’s capabilities. Along the way, you’ll also meet an incredible ecosystem of Postgres extensions like pgvector, PostGIS, pgmq, and TimescaleDB. You’ll be amazed at everything you can accomplish with Postgres! What's Inside Generative AI, geospatial, and time-series applications Modern SQL including window functions and CTEs Full-text search and JSON B-trees, GIN, GiST, HNSW, and more About the Reader For application developers, software engineers, and architects who know the basics of SQL. About the Author Denis Magda is a recognized Postgres expert and software engineer who worked on Java at Sun Microsystems and Oracle before focusing on databases and large-scale distributed systems. Quotes I was pleasantly surprised to learn many new things from this book. - From the Afterword by Vlad Mihalcea An excellent guide covering everything from basics to cutting-edge features. - Dave Cramer, PostgreSQL JDBC Maintainer Pleasant, easy to read with tonnes of great code. - Mike McQuillan, McQTech Ltd Well-organized and easy to search. - Edward Pollack, Microsoft Data Platform MVP The missing guide to understanding and using Postgres. - Mehboob Alam, POSTGRESNX, Inc.

Transform your data replication strategy into a competitive advantage with Oracle GoldenGate 23ai. This comprehensive guide delivers the practical knowledge DBAs and architects need to implement, optimize , and scale Oracle GoldenGate 23ai in production environments. Written by Oracle ACE Director Bobby Curtis, it blends deep technical expertise with real-world business insights from hundreds of implementations across manufacturing, financial services, and technology sectors. Beyond traditional replication, this book explores the groundbreaking capabilities that make GoldenGate 23ai essential for modern AI initiatives. Learn how to implement real-time vector replication for RAG systems, integrate with cloud platforms like GCP and Snowflake, and automate deployments using REST APIs and Python. Each chapter offers proven strategies to deliver measurable ROI while reducing operational risk. Whether you're upgrading from Classic GoldenGate , deploying your first cloud data pipeline, or building AI-ready data architectures, this book provides the strategic guidance and technical depth to succeed. With Bobby's signature direct approach, you'll avoid common pitfalls and implement best practices that scale with your business. What You Will Learn Master the microservices architecture and new capabilities of Oracle GoldenGate 23ai Implement secure, high-performance data replication across Oracle, PostgreSQL, and cloud databases Configure vector replication for AI and machine learning workloads, including RAG systems Design and build multi-master replication models with automatic conflict resolution Automate deployments and management using RESTful APIs and Python Optimize performance for sub-second replication lag in production environments Secure your replication environment with enterprise-grade features and compliance Upgrade from Classic to Microservices architecture with zero downtime Integrate with cloud platforms including OCI, GCP, AWS, and Azure Implement real-time data pipelines to BigQuery , Snowflake, and other cloud targets Navigate Oracle licensing models and optimize costs Who This Book Is For Database administrators, architects, and IT leaders working with Oracle GoldenGate —whether deploying for the first time, migrating from Classic architecture, or enabling AI-driven replication—will find actionable guidance on implementation, performance tuning, automation, and cloud integration. Covers unidirectional and multi-master replication and is packed with real-world use cases.

Today, host Kirk Offel sits down with Craig Pennington, CTO of Montera Infrastructure, and Joe Walsh, Montera’s Chief Delivery Officer, for an energetic, candid conversation about the future of data centers in the age of AI. With decades of combined experience across companies like PSINet, NTT, Equinix, Oracle, Digital Realty, and Facebook, Craig and Joe bring unique perspectives on how the industry has evolved — and where it’s heading next. For more about us: https://linktr.ee/overwatchmissioncritical

Explore differential testing, invariant detection, mutation analysis, and anomaly detection techniques to automatically generate comprehensive test oracles.