WEKA

💥 Epic WEKA updates coming out of NVIDIA #GTC25! 1️⃣ WEKA is integrating with the newly unveiled NVIDIA AI Data Platform to provide optimized AI infrastructure for the future of agentic AI and reasoning models. 2️⃣ We've unveiled WEKA Augmented Memory Grid— a revolutionary new capability that integrates with NVIDIA to accelerate AI inference, maximize the number of tokens processed per second, and dramatically increase token efficiency. 3️⃣ Finally, WEKA has achieved certification as a high-performance data store for NVIDIA GB200 deployments, supporting NVIDIA Cloud Partners (NCP). There’s a lot to get caught up on, so get started with a full breakdown of all the WEKA news: https://weka.ly/3XZGRC1 Yotta Data Services Private Limited | Together AI

⚡ Inference speed is capped by how fast your infrastructure can deliver tokens to your GPUs. We just open-sourced the GPUDirect Storage (GDS) integration used in WEKA’s Augmented Memory Grid — and it’s now available for vLLM and LMCache frameworks, ready for you to contribute. This is the same component we use to deliver high-throughput, low-latency token caching to eliminate redundant compute during inference. The integration connects inference servers directly to a memory-class token warehouse, unlocking time-to-first-token (TTFT) and throughput gains that legacy architectures can’t match. We’ve also launched a #WEKA-GDS-Integration channel in the vLLM Slack so you can join the conversation. See for yourself: Run your own benchmarks and learn how infrastructure purpose-built for AI can shift what’s possible. 🔗 https://weka.ly/3Fyjj1a

🚀 From 39 seconds to 2 seconds - that's what happens when AI meets the fastest storage in the cloud. New performance validation with Oracle demonstrates how the WEKA Data Platform accelerates AI workloads in the cloud. Our joint testing on OCI H100 GPU clusters achieved 269 GiB/s aggregate storage throughput and 20x faster Time to First Token performance for large language models. The breakthrough: our Augmented Memory Grid eliminates token re-computation waste by creating a high-speed token warehouse that scales to exabyte levels. At 128K context windows, Time to First Token dropped from 39 seconds to just 2 seconds - transforming AI applications from unusable to real-time responsive. This validation proves that storage performance is no longer a constraint for enterprise AI initiatives. The WEKA Data Platform delivers the speed, scale, and cost optimization that modern AI workloads demand. See the complete performance benchmarks and technical architecture: 🔗 https://weka.ly/43I6KbH

When performance, simplicity, and scale matter — customers choose WEKA. We’re excited to announce that WEKA was named a Customers’ Choice in the May 2025 Gartner Peer Insights™ “Voice of the Customer”: File and Object Storage Platforms! According to Gartner, “Vendors placed in the upper-right 'Customers’ Choice’ quadrant of the ‘Voice of the Customer’ have scores that meet or exceed the market average for both axes (User Interest and Adoption, and Overall Experience).” This recognition comes directly from enterprise IT professionals who rely on the AI-native WEKA Data Platform to accelerate time-to-insight and reduce data friction to drive real-world impact – and we’re grateful to the customers who made it possible in this recognition as a Customers’ Choice. Download the report to see what real users are saying about the WEKA experience and why they’re choosing us to power their data in the AI era. 🔗 https://weka.ly/43UDobe

🚀 We’re proud to share that WEKA was named a Customers’ Choice in the May 2025 Gartner Peer Insights™ “Voice of the Customer”: File and Object Storage Platforms! According to Gartner, “Vendors placed in the upper-right 'Customers’ Choice’ quadrant of the ‘Voice of the Customer’ have scores that meet or exceed the market average for both axes (User Interest and Adoption, and Overall Experience).” This acknowledgement is based entirely on feedback from the people who matter most — our customers – and we feel being in the Customers’ Choice quadrant acknowledges the impact the AI-native WEKA Data Platform on tackling the most demanding data challenges customers face today with speed, scale, and simplicity. Read the press release to see why customers are choosing WEKA to power their modern data infrastructures for the AI era. 🔗 https://weka.ly/4kHwkEw

🔧 The AI revolution won’t run on yesterday’s infrastructure. As models grow in size, complexity, and importance, the need for performance-optimized, AI-native data platforms is more critical than ever. That’s why WEKA is excited to sponsor World Summit AI USA in San Francisco June 18–19, helping drive the conversation among top minds in the industry on how modern infrastructure can unlock new frontiers in AI innovation. Use code WEKA20 for 20% off your registration and let’s connect at #WSAIUSA. 🔗 https://weka.ly/3SWkiv4

🚀 AI is pushing infrastructure to its limits — and WEKA is built to keep up. As models grow more complex and context windows expand, traditional architectures are struggling to deliver the performance AI workloads demand. Join us at NVIDIA #GTCParis 2025 from June 10–12, where we’ll showcase how the WEKA Data Platform integrates seamlessly with NVIDIA’s latest technologies to eliminate data bottlenecks, maximize GPU utilization, and drive next-gen AI and ML performance at scale. If your infrastructure is holding back your AI initiatives, this is the conversation for you. Schedule a meeting and let’s get started. 🔗 https://weka.ly/45zvOnN

🧩 Scaling LLM inference is exposing critical limitations in legacy infrastructure. As context windows expand into the millions of tokens and inference pipelines grow increasingly memory-bound, traditional architectures — especially those relying on local GPU memory for KV cache — are becoming performance and cost bottlenecks. WEKA’s Augmented Memory Grid was built to address this challenge head-on. It’s a memory-class token warehouse designed to support massively parallel inference workloads with remote, low-latency, high-throughput access — achieving up to 41x faster time to first token (TTFT) in real-world testing. WEKA’s Callan Fox unpacks the architecture behind this breakthrough and how it’s shifting the performance ceiling for LLM inference while reshaping the economics of AI infrastructure. 🔗 https://weka.ly/43uMDO1

📐 Planning an NVIDIA DGX SuperPOD deployment? Don’t let storage drag down your performance — or your budget. AI infrastructure at scale demands more than just speed. It needs a data platform that can match GPU performance while delivering economic efficiency. Traditional storage wasn’t built for this. That’s why we created the WEKA TCO Calculator, helping you assess WEKApod’s potential total cost of ownership advantage early in your planning process. Build smarter from day one. 🔗 https://weka.ly/3Z45e2f

🚀 WEKA is growing fast — and we’re hiring! We’ve got two huge opportunities to join the sales team shaping the future of AI infrastructure: Regional Sales Manager, U.S. West will lead enterprise growth across the western U.S. and bring cutting-edge AI infrastructure solutions to new customers: https://weka.ly/4k6HyTc Account Executive, Strategic Pursuits will help land and grow high-impact customer relationships and ensure successful deployment of the WEKA Data Platform: https://weka.ly/3FB9Iqi We're looking for strong sellers and relationship builders who want to go big. Know someone ready to level up and crush it in a high-growth environment? Let’s talk. 🔗 https://weka.ly/4kppDHm