Rick Grinnell

ROCm 7 enables engineering teams to deploy new AI models immediately, with optimized Triton kernels, dynamic batching, and vendor-agnostic scaling across GPUs. To explore how ROCm 7.0 supports inference and fine-tuning at scale or deploy it in your environment check out this blog: https://bit.ly/4mGgWZF

352

10 Comments

See what’s new in Codeium! Join us tomorrow, Feb 19 at 11am CST / 9am PST for a live demo! Our engineers Alexander Baker and Abhay A. will showcase the latest enterprise updates, including: ✨ Windsurf Enhancements – Streamline development with our most advanced editor updates. 🤖 New Features – Autogenerated Memories, Web Search, Image Upload, and more. 🔧 Developer Tools – Devcontainer improvements, Cascade Memories, and new integrations. 📅 Roadmap & Company Updates – Get a sneak peek at what’s coming next. If you're a developer, engineering leader, or tech enthusiast, you won’t want to miss this deep dive into Codeium's latest capabilities. 👉 Register now: https://lu.ma/2oe0l3nz

75

In a new report, IAPS researchers Erich Grunewald and Asher Brass Gershovich look at the security of AI data centers housing frontier AI systems, which are potential targets for sophisticated adversaries like China and Russia seeking to steal IP or sabotage infrastructure. AI data center construction is growing rapidly. US companies increased capex by 40% from 2022-24, with hundreds of billions planned yearly. But these facilities aren't being built to withstand nation-state threats, and security measures are difficult or impossible to retrofit. The threats are substantial. Chinese cyber operations have demonstrated the ability to penetrate American networks and remain undetected for months or years. AI and data center companies may underinvest in security since they do not bear the full societal costs of breaches. Efforts are needed across the board, but some neglected attacker techniques especially need attention: ➡️ Side-channel attacks ➡️ Hardware supply chain attacks ➡️ Methods for exfiltrating large AI model weights The report makes 4 key recommendations: 1️⃣ Develop an AI data center security standard 2️⃣ Fund and incentivize key R&D projects 3️⃣ Establish cyber incident intel sharing between AI companies and authorities 4️⃣ Map critical component supply chains and shift them out of China

92

5 Comments

Hedera is designed for network stability, fair access, and efficient transaction processing. One key mechanism enabling this is throttling, which regulates the rate of operations to prevent overload and ensure seamless performance. In this blog, we provide a comprehensive technical overview of Hedera's throttles, explaining their purpose, functionality, and configuration. Read now: https://lnkd.in/dntfvWss

82

Anna Horner joins Caolán McCarron, Director of Product Management, to discuss how AtlasVision is setting a new standard in trading infrastructure monitoring and market data transparency.   AtlasVision is a cutting-edge monitoring platform designed for real-time diagnostics and alerting across market data feeds and circuit connectivity. It gives Options' clients the visibility they need to quickly detect latency issues, identify root causes, and maintain a competitive edge.   Through a dedicated, user-friendly portal with single sign-on, clients gain full transparency into the infrastructure that supports their market data, with features including: - Real-time outage and latency notifications - Rapid root-cause identification - Granular detection of market data gaps - Seamless API integration with internal tools With the recent full integration with Packets2Disk, clients can now access deep, packet-level analytics for market data, trading, and network insights. Visit https://lnkd.in/eu2MSMP3 for more information.   #AtlasVision #MarketData #NetworkTransparency #InfrastructureInsights #TradingTechnology

157

2 Comments

Edge AI Surveillance Deployment: Key Technical Considerations DFI recently published an insightful article examining the practical challenges of deploying intelligent video surveillance at scale, featuring DEEPX's AI acceleration technology as a key solution component. With the market projected to grow from $14.40B to $21.65B by 2030, the article identifies four critical factors that determine deployment success: 1. Performance & Latency Edge processing eliminates cloud delays while addressing privacy and bandwidth constraints for real-time decision-making. 2. Environmental Resilience Industrial deployments require -40°C to 85°C operation with shock/vibration resistance—eliminating most commercial hardware. 3. System Integration Unified platforms supporting diverse AI models across heterogeneous hardware reduce engineering complexity. 4. Scalability & Maintenance Efficient orchestration tools for managing model updates and system health across expanding deployments. DFI's Solution Approach: DFI demonstrates how these challenges can be overcome through DEEPX's ultra-efficient AI acceleration technology. The article showcases their EC700-ASL system integrated with DEEPX's DX-M1 chip and Network Optix software, delivering impressive results: 4-channel object detection processing while maintaining just 25% CPU utilization. This shows how optimized AI acceleration enables multiple workloads in resource-constrained environments. Smart surveillance deployments powered by industrial-grade hardware and ultra-efficient AI semiconductors will enable more sophisticated applications across public safety, transportation, and industrial monitoring in a cost-efficient manner. Read DFI's full technical analysis: https://lnkd.in/gRmvY7iz #EdgeAI #AISurveillance #DFI #DEEPX

54

SIERRA Partner, Ashish Kakran, joined an insightful panel at the Confidential Computing Summit, hosted by OPAQUE Systems and the Confidential Computing Consortium. The conversation spanned some of the most urgent challenges and opportunities in enterprise AI and security: • Shadow AI is becoming the new shadow IT—introducing risk, sprawl, and data leakage across the enterprise. • Data readiness remains a gating factor for GenAI adoption. Startups helping customers organize and govern their data are critical enablers. • The AI attack surface is growing fast. As agents and LLMs enter enterprise workflows, security and compliance must be rethought from the ground up. • And most notably: AI agents are coming—and they’re already reshaping how teams work. This shift opens up whitespace around identity, observability, orchestration, and beyond. Thanks to the organizers and fellow panelists for pushing the conversation forward. If you're building at the intersection of AI, security, and infrastructure, we’d love to connect. #AI #ConfidentialComputing #EnterpriseSecurity #Startups #SierraVentures

46

1 Comment

How did Crusoe get their LLM #cloud offerings right from the start? By refusing to leave performance to chance. They used Keysight CyPerf to rigorously test, optimize, and scale their AI/#ML infrastructure. Discover why a testing-first approach outperforms trial and error when it comes to AI/ML at scale: https://ow.ly/8oUm50VHjva

55

🎧 Data Management Frontiers: Navigating the Semiconductor Landscape with Joe Addiego of Brave Capital 🎧 Our latest episode of #AdvantestTalksSemi is now live! Join us as we delve into the world of data management with Joe Addiego, exploring how advancements in real-time databases and NoSQL systems are shaping the future of semiconductor technologies. This episode reveals how these innovations drive the semiconductor industry towards faster, more efficient, and scalable data solutions. 🎙️ Featuring Expert Insights: • Joe Addiego, veteran tech investor, discusses the transformative impact of data management technologies from punch cards to AI-driven systems, and their critical roles in modern semiconductor processes. • Hosted by Don (Joan Hong) Ong, Director and Head of Innovation for the Advantest Field Service Business Group, steering the conversation on the future of data strategies in high-tech industries. 🎙️ Key Takeaways: • Data-Driven Decisions: Understand how the integration of advanced data management systems enhances decision-making processes in semiconductor manufacturing. • Technological Milestones: Discover the evolution from traditional database systems to sophisticated NoSQL and real-time databases that handle vast amounts of data with unprecedented speed and efficiency. • Future of Data Management: Explore the potential developments in data infrastructure and how they're poised to further revolutionize the semiconductor sector. Tune in now to gain invaluable insights into the cutting-edge of data management in the semiconductor industry! https://lnkd.in/gDpCciqF #AdvantestTalksSemi #DataManagement #BestTechPodcast

58

CVector has raised $1.5M in pre-seed funding led by Schematic Ventures to scale its AI-driven platform for industrial optimization. Co-founded by Richard Zhang and Tyler Ruggles, the startup integrates real-time data across legacy systems to support sectors like chemicals, energy, and utilities—bringing next-gen intelligence to uptime-critical infrastructure. https://lnkd.in/evu__KG3 #IndustrialAI #AIInfrastructure #ManufacturingTech #EnergyTech #AIOptimization #PreSeedFunding Alissa Ganter Alex Kamath Emilie Ogden-Fung Joshua Napoli Samuel Corwin

69

6 Comments

From zkVMs to TLSNotary to ZKP2P, platforms are leveraging zero knowledge cryptography and succinctness to give developers across industries new affordances with which to build software. On August 15, 2025, Tower Research Ventures will sit down with Richard Z of ZKP2P to talk about his project and the evolving ZK tooling landscape. Open to cryptographers, researchers, and entrepreneurs exploring applied cryptography! https://lu.ma/ca3hsfpt

44

1 Comment

IonQ continues to make the rounds at #SXSW in Austin, meeting industry leaders and sharing more about the potential of quantum computing. Yesterday, IonQ’s Chief Revenue Officer, Rima Alameddine, joined TreQ’s Mandy Birch, IBM’s Charles Chung, and Vizias’ Jay Boisseau to discuss the path to mainstream quantum computing. The panel addressed the problems quantum can solve for businesses and the challenges of scaling quantum technologies. Rima touched on the time and knowledge it takes to develop quantum algorithms, underscoring why it’s important for businesses to invest in talent and expertise now if they want to lead in this emerging field. IonQ also hosted a social event at Remedy, bringing together quantum innovators, enthusiasts, and industry leaders to network and exchange ideas. Special thanks to Jay Boisseau for helping coordinate and to Texas Rep Linda Garcia, Google’s William Magro, Q-CTRL’s Michael Biercuk, Strangeworks’ William Hurley, and UK government VC Meredith Fisher for stopping by! Dina Moy Philip Farah Anuja Kochar Singh David Wiese

94

1 Comment