Microsoft has introduced Cloud Key Management System (KMS) signing support for Microsoft's Cryptography API-Next, allowing developers to digitally sign their code and attest to its legitimacy. This feature is designed to build trust in the software world by providing a secure way to generate, manage, and control access to cryptographic keys. Google Cloud's Cloud KMS offers a user-friendly interface for creating, storing, and performing cryptographic operations such as code signing with keys in tamper-resistant Cloud hardware security modules (Cloud HSM).

Source: https://cloud.google.com/blog/

RAG design and LlamaIndex, a powerful framework for building RAG applications, simplifies the process of connecting to data sources, structuring information, and querying with LLMs. It breaks down the RAG workflow into indexing and storage, retrieval, node post-processing, and response synthesis. LlamaIndex offers various techniques and integrations to achieve this.

Source: https://cloud.google.com/blog/

To adapt the Gemini Large Language Model (LLM) for specific use cases, a task-specific annotated dataset is used for fine-tuning. This dataset includes input examples and their corresponding desired outputs, allowing the model to learn a specific task. Two common approaches to supervised fine-tuning are full fine-tuning, which updates all model parameters, and partial fine-tuning, which updates only a subset of parameters.

Source: https://cloud.google.com/blog/

Source: https://cloud.google.com/blog/

Source: https://cloud.google.com/blog/

Source: https://cloud.google.com/blog/

The development of a customized dual encoder model for a production retrieval system is a challenging task that requires data science and ML engineering experience. To solve this problem, Advanced RAG approaches using Large Language Models (LLM) have emerged. These approaches use LLM reasoning capabilities to generate possible answer texts and expand the original query with possible answer candidates. However, this method adds LLM prediction latency and high cost for every query.

Source: https://cloud.google.com/blog/

Google Cloud is committed to ensuring data safety and security through Confidential Computing, a hardware-based Trusted Execution Environment (TEE) that protects data while it's being used and processed. The company has been an early adopter and investor in Confidential Computing products and solutions for over four years, growing its offerings and adding new capabilities.

Source: https://cloud.google.com/blog/

Google, a pioneer in AI, has been leveraging the technology to organize the world's information and bring commercial-grade data analytics, security, and gen AI to the public sector. The State of Illinois has deployed BEACON, a Google AI-based platform, to match families with mental health support resources for their children. The Defense Innovation Unit (DIU) has partnered with Google Cloud to accelerate their mission.

Source: https://cloud.google.com/blog/

Google Cloud is partnering with Aiven to bring AlloyDB Omni, a high-performance, 100% PostgreSQL-compatible downloadable database, to Aiven's platform. AlloyDB Omni features a scalable online transaction processing architecture, near real-time analytics, and vector processing. This partnership expands AlloyDB Omni's reach and availability, offering developers and businesses more flexibility in deploying and managing their PostgreSQL workloads.

Source: https://cloud.google.com/blog/