Duality Technologies Unleashes Secure GenAI Workflows on NVIDIA GPUs for Enhanced AI Performance

Duality Technologies Enhances GenAI Workflows on NVIDIA GPUs



Duality Technologies, a trailblazer in the realm of privacy-enhancing technologies and secure data collaboration, has made waves in the tech industry with its recent announcement. This latest offering, which incorporates support for Google Cloud's Confidential Computing portfolio, allows organizations to leverage NVIDIA GPU-powered confidential virtual machines. This is a significant step that enables enterprises to execute large-scale AI workloads securely, encompassing essential tasks like training and inference for large language models (LLMs).

With the introduction of this new capability, the Duality Platform now accommodates GPU-backed large language model inference and encrypted Retrieval-Augmented Generation (RAG) in trusted execution environments (TEEs). This enhancement marks a monumental shift from the previous system that relied solely on CPU, thereby significantly boosting performance and efficiency.

Dr. Alon Kaufman, CEO and Co-Founder of Duality Technologies, expressed excitement about this development: "This changes the game for our clients. They now have the opportunity to implement privacy-preserving AI with LLMs on a production scale. The hurdles posed by performance bottlenecks in secure computing are now eliminated, making secure LLM training and inference feasible and practical."

One of the key features of this launch is its integration with Google Cloud's Confidential Space and the confidential NVIDIA H100-powered virtual machines. Coupled with support for Intel TDX and Cloud KMS integration, customers can rest assured that their data remains protected across every step of the AI workflow.

The platform has already demonstrated its effectiveness, successfully validating the operation of a Mistral-7B model using encrypted vector RAG through Faiss in a fully confidential pipeline. Nelly Porter, Director of Product Management at Google Cloud, also shared insights about this innovation, stating, "With Confidential GPUs, organizations can manage sensitive AI workloads entirely within trusted execution environments, all while maintaining high performance levels. The combination of NVIDIA H100-powered confidential VMs with Duality’s encrypted workflows facilitates larger-scale LLM training and inference, with comprehensive protection against data leakage."

Key Highlights of the Launch


1. GPU Support for Confidential AI: Now organizations can securely run LLMs and encrypted RAG on Confidential NVIDIA H100s.
2. Scalable Performance: Users will observe runtimes that are orders of magnitude faster compared to CPU-only workloads, making this solution highly efficient.
3. Enterprise-Ready Solutions: The offering meets the stringent requirements of regulated industries, including defense, healthcare, and companies focused on AI.
4. Seamless Cloud Integration: The capability is easily accessible through the Dynamic Workload Scheduler integrated within Google Cloud’s Confidential Space.

Previously, the landscape of confidential AI had its limitations, constrained to CPU-only environments that were adequate for basic testing but fell short of the demands required for extensive AI functionalities. The introduction of Confidential GPUs within the confidential computing portfolio transforms this dynamic, allowing Duality’s clients to securely conduct both LLM training and inference within Trusted Execution Environments. Consequently, this breakthrough facilitates high-throughput and privacy-preserving AI workloads, effectively unlocking a spectrum of new applications across various industries.

The initial rollout of this advanced capability is currently available on the Google Cloud Confidential A3 virtual machine type, with a broader rollout anticipated later in the year. Organizations looking to explore this groundbreaking offering should visit dualitytech.com for further information.

About Duality Technologies


Duality Technologies stands at the forefront of privacy-enhancing technologies (PETs) and aims to foster secure AI collaboration. Their innovations empower regulated industries and governments to utilize sensitive data across organizational and jurisdictional constraints. Potential clients can consult with Duality's offerings to harness the power of data while ensuring privacy and security.

For media inquiries, please reach out to:
Diane McKaye
Si14 Global Communications
[email protected]
+44 7771 926726

Topics Business Technology)

【About Using Articles】

You can freely use the title and article content by linking to the page where the article is posted.
※ Images cannot be used.

【About Links】

Links are free to use.