Gurpratap Virdi

✨ From Magnifying Glass to Machine Learning ✨ AI for Toy Cataloging and Valuations At Redmond’s Forge, we’ve lived through the shift from manual toy valuation to an AI-powered approach. On the left: painstakingly flipping through reference books, jotting notes, and hoping to find comparable LEGO minifigures. On the right: AI tools instantly analyzing sealed vs. used values, highlighting museum-worthy sets, and even recommending where in the LEGO City, build a MOC, or the best resale platforms. This isn’t about replacing tradition, it’s about enhancing craftsmanship with technology. We can still respect the history of collecting while embracing the future of AI-assisted valuation. 🔗 Read the full post: https://lnkd.in/ebpi4Bcv #AI #GenerativeAI #LEGO #Collectibles #RedmondsForge #FutureOfWork #ArtificialIntelligence

4

Curious to see how the MI355X has benchmarked in testing so far? Signal65 in collaboration with AMD ran real world workload benchmarks here: https://lnkd.in/dYv-MMGH AMD Instinct MI355X-Examining Next-Generation Enterprise AI Performance - Signal65

13

3 Comments

Nice write-up! Some thoughts. (1) Karma pays back - the Rivos team really paid into the ecosystem with RISE, RVI Server SoC WG, to name a few. Many of the folks left big gigs to take on a high risk path, but did this while strengthening the ecosystem cause for others too. Sometimes big risk makes you inward focused, they were not. Huge congratulations to the Rivos Inc. team and leadership. (2) Using RISC-V to solve problems, it's not an ISA vs ISA competition - Rivos' software-first mindset helped as they honed in on a big customers big problem that needed solving, using the RISC-V hw/sw platform to tailor their compute engines for the task at hand - GPU/SIMD, CPU. At Akeana, we have been hearing this need from AI chip customers and discussed task-specific cores in this webinar (https://lnkd.in/gd4XaZpn).  (3) RISC-V as a hw/sw co-design platform - any company planning on building chips for the future decades will eventually have an in-house RISC-V platform/team. Meta has clearly seen this signal and acted. Nvidia had shown the light in their "one architecture" keynote at the '24 RVI summit (https://lnkd.in/gHvH4_G4). BTW, the next RVI NA summit is in Santa Clara this month (Oct 22-23, https://lnkd.in/gAt8W_92).

12

Roksan quietly made a big move with these new Caspian 4G separates. For anyone who’s outgrown an integrated amp but doesn’t want to abandon modern streaming, this combo makes a lot of sense. BluOS, proper phono support, fully balanced architecture, and a power amp that can scale all the way to serious monoblock output. It’s refreshing to see a brand focus on doing the fundamentals right instead of chasing spec-sheet gimmicks. Definitely worth paying attention to if you’re building a long-term system. #Audiophile #HiFi #Roksan #TwoChannel #HighEndAudio

2

Great AI products are only as good as the data powering them. This article perfectly articulates why Gracenote structured entertainment data and "System of Record" IDs are the essential substrate for the CTV ecosystem—powering both Search & Discovery and Ad systems. > By acting as the bridge between LLMs and authoritative datasets, our MCP Server turns probabilistic AI outputs into grounded, reliable results. > Gracenote Content Connect enables CTV ad buys with a much deeper understanding of the content, with a standard industry taxonomy. We’re not just providing data; we’re building the products that define the future of CTV consumption and ad buying. 🚀 Check out the full story: https://lnkd.in/g-PDbWE3

47

So proud of our cross-functional team for delivering AI Overviews and Ask Gemini in Drive! These features represent a huge step forward in how we help users navigate and synthesize their information. Incredible work by everyone involved! More details: https://lnkd.in/eKsPpgbg

25

Like most in my network, I spent most of my career in traditional Ethernet (front-end networks). With AI, there are three new networks are emerging - Scale up: GPU-to-GPU Scale out: Node-to-Node Scale across: DC-to-DC   The following 3 white papers offer a good summary of “why”, “what”, and “where do we go from here” for AI backend scale-out networks: 1. Datacenter Ethernet and RDMA: Issues at Hyperscale: AI workloads require lossless networks because they depend on extreme levels of parallelism and synchronization.  The industry adopted RDMA (Remote Direct Memory Access) over Converged Ethernet (RoCE) as a protocol for AI backend scale-out networks. This paper discusses key challenges and inefficiencies with the RoCE standard. Link: https://lnkd.in/gKqSb3XB 2. I’ve Got 99 Problems But FLOPS Ain’t One: As AI Infrastructure deployments is a core competitive advantage, Hyperscalers tend to be secretive about the challenges they face (and the solutions derived to address the problems) in building global-scale network infrastructures. The authors of this white paper used a research-based approach to model network challenges in supporting mega clusters and identified potential research areas for further exploration. Link: https://lnkd.in/g_V59r2A 3. Ultra Ethernet’s Design Principles and Architectural Innovations: The 562-page-long Ultra Ethernet Specification 1.0 is intimidating. Ultra Ethernet specification 1.0 was designed to address AI and HPC workload network requirements. Key guiding principles include: Massive Scalability, High Performance, Compatibility and Vendor Differentiation. This white paper provides an easily digestible summary of key advancements proposed in Ultra Ethernet specification. Link: https://lnkd.in/gBFr-Vtv Ultra Ethernet Specification: Ultra Ethernet Specification v1.0 June 11, 2025: https://lnkd.in/gTmX73rV I would be remiss if I did not mention Torsten Hoefler 🇨🇭 who not only made contributions to the specification, but is also democratizing the specification. Thank you, Torsten!  Torsten Hoefler Talk @ 2025 Swiss Conference: Next Generation AI and HPC Networking with Ultra Ethernet: Torsten Hoefler Talk: https://lnkd.in/gc9kZYB8 Are there other AI Backend Scale Out resources that you came across? #AINetworking #UEC #ScaleOutNetwork

178

6 Comments

Midmarket customers need simplicity, not hyperscale complexity. 🧩 While AWS, Azure, and Google Cloud were built for global scale, they often fall short for businesses that value clarity, control, and real support. This white paper breaks down the key challenges—like opaque pricing, overengineering, and vendor lock-in—and shows why more partners are choosing CaaB. ✅ Transparent pricing ✅ Turnkey disaster recovery ✅ No lock-in Download the white paper to explore how a purpose-built platform like CaaB - Cloud as a Business helps reduce risk, not shift it.

1

1 Comment