#USA #San Jose #NVIDIA #AI Data Platform #WEKA

WEKA Augments AI Performance at GTC 2025

At the GTC 2025 event, WEKA, recognized for its AI-native data platform, took a substantial step forward by integrating with the NVIDIA AI Data Platform reference design. This collaboration aims to enhance AI infrastructure tailored for the future of agentic AI and advanced reasoning models. Featuring newly achieved storage certifications for NVIDIA Cloud partners, WEKA is establishing a robust baseline for enterprises to leverage AI's full potential.

Introducing the Augmented Memory Grid

A standout highlight from WEKA was the announcement of the Augmented Memory Grid™ capability. This groundbreaking feature merges WEKA Data Platform software with NVIDIA's advanced computing and enterprise software. The objective is clear: to expedite AI inference processes, elevate the number of tokens processed per second, and significantly boost token efficiency.

Reshaping Data Infrastructure

The NVIDIA AI Data Platform is redefining how enterprises structure their data for agentic AI applications. By providing a customizable design integrating NVIDIA's Blackwell platform, BlueField DPUs, Spectrum-X networking, and AI Enterprise software, businesses can transform their data into actionable insights. The synergy between WEKA and NVIDIA allows organizations to benefit from this platform, creating scalable and high-performance foundations for enterprise-centric AI, thereby enhancing AI model performance and accuracy.

Overcoming Memory Limitations

As AI agents increasingly take on complex decision-making and problem-solving tasks, the demand for robust infrastructure has never been higher. WEKA's Augmented Memory Grid addresses this challenge by offering the capability to extend model inference memory, allowing for the processing of large-scale data with easier petabyte capacity. This innovation promises near-memory speed performance with microsecond latencies, ensuring swift token processing and unparalleled reasoning outcomes.

Key Benefits

• - Faster Token Processing: For instance, during tests with 105,000 tokens, the Augmented Memory Grid demonstrated a 41-fold reduction in the time it took to retrieve the first token, exemplifying efficiency in AI inference.
• - Cost Efficiency: Inferencing clusters can achieve greater token throughput, reducing token processing costs across the entire inference framework by up to 24%.

Achievements in Certifications

WEKA's flagship product, the WEKApod™ Nitro Data Platform, has garnered certification as a top-tier storage solution for NVIDIA Cloud Partner deployments. By aligning with NVIDIA's standards for AI development, WEKApod serves as a high-performance data solution that supports NVIDIA's powerful GPU technologies.
WEKApod has also received the NVIDIA-Certified Systems Storage designation, affirming its compatibility with NVIDIA's best practices to optimize storage performance for diverse enterprise AI workloads.

Advancing AI Storage Solutions

WEKA continues to collaborate closely with NVIDIA, providing essential high-performance storage solutions to organizations leveraging the NVIDIA AI Data Platform. The strategic focus for WEKA is to eliminate data challenges constraining AI innovation. As Nilesh Patel, chief product officer at WEKA, emphasized, combining WEKA's technological advancements with NVIDIA's capabilities is akin to breaking the sound barrier for aerospace innovation, shattering the limitations in AI memory and enhancing token efficiency across the board.

Industry Reception

Experts from notable companies have voiced their enthusiasm about this technological advancement. For instance, Ce Zhang, CTO at Together AI, expressed excitement for utilizing WEKA's new capability, highlighting improved latency and efficient data caching for their inference engine.
Similarly, Sunil Gupta, the co-founder and CEO of Yotta Data Services, noted the pivotal role WEKA's platform plays in amplifying performance for India's leading AI supercomputing platform.

Availability

WEKA's NCP reference architecture for NVIDIA Blackwell systems is set to launch later this month. The Augmented Memory Grid feature for WEKA Data Platform will be generally available in Spring 2025, expected to deliver significant improvements in AI model efficiency and capacity.

In summary, WEKA's advancements at GTC 2025 highlight a remarkable leap in the integration of AI capabilities with cutting-edge hardware solutions. By addressing memory barriers and optimizing processing times, WEKA stands at the forefront of the AI infrastructure revolution, empowering businesses to redefine their operational efficiencies with AI.