proteanTecs’ Post

📅 TOMORROW! WEBINAR & LIVE DEMOS: Accelerating NPI with Deep Data: From First Silicon to Volume 👉 https://hubs.la/Q03Ss-bN0 NPI (New Product Introduction) is where innovation meets production, the moment of truth for every new chip. The goal? Ensure your silicon meets specs and can ramp to volume with confidence. But at advanced technology nodes, NPI is becoming tougher than ever. Shrinking process margins, complex architectures, and exploding data demands are raising the bar for success. Common NPI challenges include: 🔸 Wider performance spreads caused by manufacturing variability at nanoscale geometries 🔸 Escalating test time, cost, and data management complexity 🔸 Reliability that’s harder to predict as aging and stress effects intensify 🔸 Faster time-to-market pressure colliding with slower model-to-silicon feedback Discover how on-chip parametric deep data can help you overcome these challenges, enabling faster silicon learning, better yield, and smoother ramp to volume. Join us tomorrow for a webinar profiling real-world use cases and live demos. 👉 https://hubs.la/Q03Ss-bN0

Find out how to boost shop productivity and make smarter machining decisions with digital tools and artificial intelligence at the Modern Machine Shop’s Top Shops Conference 2025 (TS25). Our Vice President of Engineering, Ed Rusnica, will talk about how to “Improve Shop Efficiency Using AI and Digital Machining Tools” on November 11. He’ll share how our AI-driven plug-ins and partnership with Toolpath Labs can help optimize tool selection, refine toolpath strategies and make application engineering more accessible. Learn more about the event: https://heyor.ca/x2OT25

During the Semicon West few weeks ago in Phoenix, I had the pleasure to discuss about the importance of hybrid Cu bonding in enabling continued scaling of heterogenous integration of chiplets. Applied Materials launched Industry’s first integrated Die-to-Wafer bonding system, Kinex, in close collaboration with industry-leading Besi bonder, ushering a new wave of chiplet integration capability with significantly higher bandwidth and lower power consumption for next gen AI accelerators. My talk covered the key capabilities of Kinex D2W bonding system which enables high quality, highly yielding hybrid Cu bonding in a high volume manufacturing environment.

Great two days in Munich at the 2nd Constructing Semiconductor FAB Summit Europe by Innovatrix International. The event was an excellent opportunity to connect and exchange ideas with industry leading experts. Lots of inspiring discussions about best practices in fab design and construction. Key takeaways: 📊Digital twins and data are becoming the new norm with several practical AI solutions already in use. 🏭Whether gigafactory scale or modular small fab the fundamentals stay the same: safety first, contamination control and quality with strong in-line measurement. 👷🏻♂️Gained valuable insights into managing fab building projects and facility systems end to end. Lots of inspiration to take back to our own fab expansion project at Canatu.

Last week in Washington, D.C. I attended the SEMAFOR World Economy Summit, sitting for a fireside chat with tech editor, Reed Albergotti. Our discussion focused on the rise of AI and how it's driving unprecedented design complexity, cost pressure and an accelerated pace of innovation for engineering teams that design silicon-powered, software-defined and AI-enhanced products.   To tame these challenges, we must re-engineer how those products are engineered, from the compute, to the engines and solvers, and to the workflows. Product R&D teams need holistic solutions with a deeper integration of electronics and physics — like those available from Synopsys Inc — to design, optimize and virtualize the silicon and the entire system.   The rise of AI means we must also see a rise in systems-level thinking from the future engineering workforce, versus domain-specific expertise. These significant shifts underscore the need for even greater collaboration between industry and academia to ensure that emerging engineers are thoroughly skilled in AI.   Check out our full conversation here: https://bit.ly/48IueBF

We’re witnessing a transition in how engineering simulation is applied: from specialist tools used during design to intuitive apps powering decisions across business functions. The article, by Björn Sjödín at COMSOL, explains how these simulation apps let frontline engineers, operations teams, and business strategists ask “What if?” and get answers fast—turning models into accessible decision engines. If your team is looking to scale insights, drive agility, or break down silos between engineering and business, this read is timely. Read more: https://lnkd.in/eUC6Sauv

I just got back from #INFORMS2025, where I had the opportunity to present my latest research on individualized mixed-model assemblies. While customers expect products to be individualized, manufacturers are increasingly pressured to produce locally. This leads to production that is characterized by high workload variance between and within products. Using data from a German manufacturer, we presented: - a new approach for precedence graphs that accurately captures product individualization through different kinds of options, - T-Bars as a method for straightforward, optimized line balancing despite workload spreads at stations, and - two approaches for building bespoke takt time groups to increase line utilization, one suited for a make-to-order environment and one for an assemble-to-order environment. We showcased the implications of both approaches and defined the conditions under which each is favorable. Presenting at INFORMS was a fantastic opportunity to exchange ideas with scholars and practitioners tackling similar challenges in modern manufacturing.

This will be a good one! ikeGPS’s new automation capabilities make analyzing distribution assets much less time and resource intensive.

A quite AD-relevant Monday at #ICCV2025 is over: Henrik Arnelid and myself were treated to workshops, poster sessions and discussions with companies such as Nvidia, Zoox, Tesla, Kodiak, Wayve ... Some topics that I found it cool to learn about: • Leveraging VLMs for clip curation/triaging and description of edge case 'counterfactuals' to re-balance training datasets in order to mitigate undesired behavior • Using latent space comparison to find out what makes different logs 'similar' in non-obvious ways; enabling higher-efficiency simulation strategies (downsampling the dreaded 'infinite test space' by identifying unexpected overlaps) • Tapping into reasoning composites (3D occupancy, segmentation, language) to improve causal mapping between pixels and actions in otherwise opaque E2E models • Benchmarking (poor) spatial reasoning performance between different VLMs, and the prospect of building models more capable of this based on the learnings • Generating synthetic 3D radar data - informed and verified by real-world recordings - by means of gaussian splatting • Closing feedback loops between (systems-level) AD requirements or QC pass/fail reports and related data needs with the help of LLMs/VLMs ICCV is a research-heavy conference, and many people come with more questions than answers. After my first day here, one thing seems certain, though: AD players definitely have an appetite for next-gen data paradigms. This space is not getting boring for Kognic anytime soon!

SNIA and and our partners will be at SC25 showcasing open standards for #AI and #HPC at SC25. Stop by the Open Standards Pavilion (booth 211) for: - AI and HPC Demonstrations: Hands-on exhibits at the booth will feature SNIA Swordfish® scalable storage management and 24G+ SAS technologies; AI and HPC solutions for persistent and CXL® memory programming and memory-to-memory data movement; time and performance-based Total Cost of Ownership (TCO) models; and new E2 SSD form factor standards. - Featured Communities: SNIA will showcase its Storage Management, Compute, Memory and Storage, and SCSI Trade Association (STA) Communities. - StorageAI™: SNIA team members at the booth will discuss their collaborative open standards StorageAI project to address AI workload challenges, including data pipeline inefficiencies, idle GPUs, memory tiering, data movement, and system latency.https://https://lnkd.in/en_zcW8W We hope to see you! SC Conference Series