Alan Nichol

🔬 Meet our researchers 🔬 Science doesn’t slow down – but your code might. That’s where today's HAFIS project steps in to bridge the gap between scientific code and the kind of computing it deserves. Py2HPC by Claudia Comito from Forschungszentrum Jülich and Hassan Nassar from Karlsruher Institut für Technologie (KIT). 🔹 Project focus Py2HPC project aims to bridge this gap by fostering collaboration between domain scientists and HPC experts. Py2HPC provides a structured environment and a growing library ecosystem where scientists are supported by specialists who help them adapt and accelerate their Python-based applications using the latest HPC tools and platforms. 🔹 Vision & Goals Central to the project is the development of reusable software components and templates that map well to HPC architectures, including multi-node clusters, GPUs, and FPGAs. HPC experts assist scientists in identifying critical computation paths, choosing suitable parallelization strategies, and deploying optimized versions of their algorithms on the appropriate hardware. This human-in-the-loop approach ensures both accessibility and performance, while reducing the steep learning curve typically associated with hardware acceleration. 🔹 Challenges to overcome The motivation for this project is the increasing reliance on Python in scientific computing and the simultaneous rise of heterogeneous high-performance computing (HPC) platforms, such as GPUs and FPGAs. While Python remains the dominant language for prototyping and experimentation, it often fails to utilize the full potential of modern accelerators due to performance bottlenecks and limited low-level hardware integration. Many scientists, especially those without formal training in HPC or hardware programming, struggle to scale their solutions effectively. 🔹 Most valued feedback Realizing that identifying our MVP early and investing in our internal structure is just as critical as showcasing our technical advantage. This shifted how we prioritize and plan. 🔹 Why HAFIS matters Connecting with like-minded innovators has helped us see our idea's potential and gave us the confidence to communicate its value clearly. Thank you Claudia and Hassan, for making brilliant ideas run more efficiently! 💻 What would you like to know about Py2HPC? Leave your question in the comments! #Phython #Programming #HAFIS

5

We're excited to be part of the Helmholtz Foundation Model Initiative (HFMI), a major step toward harnessing the power of AI for science. Helmholtz Imaging is contributing to two of the seven funded projects, as well as the Synergy Unit: ◾ The Human Radiome Project – aiming to drive a paradigm shift in medical research by providing novel insights into human health and disease through the power of AI ◾ AqQua – using AI and high-performance computing (HPC) to analyze large collections of plankton images from oceans around the world ◾ The Synergy Unit – supporting all HFMI projects by developing general principles for AI model creation and use across scientific domains HFMI provides a unique framework for developing AI-powered research tools, built within the Helmholtz Association but made available to the global research community. We're proud to help shape the future of data-driven science. 👉 Learn more about HFMI and all funded projects in the post below. 👉 For a closer look at the projects Helmholtz Imaging is involved in, visit our news page: https://lnkd.in/eutdGJTx #30Jahre30Stories #Helmholtz #FoundationModels #AI #OpenScience #HFMI #HelmholtzImaging

7

📢 📢 📢  A Radar Echo Simulator for the Synthesis of Randomized Training Data Sets in the Context of AI-Based Applications 🧑‍🔬 Jonas Schorlemer, Jochen Altholz, Jan Barowski, Christoph Baer, Ilona Rolfes, and Christian Schulz 🏫 Ruhr University Bochum 💥 Supervised machine learning algorithms usually require huge labeled data sets to produce sufficiently good results. For many applications, these data sets are still not available today, and the reasons for this can be manifold. As a solution, the missing training data can be generated by fast simulators. This procedure is well studied and allows filling possible gaps in the training data, which can further improve the results of a machine learning model. For this reason, this article deals with the development of a two-dimensional electromagnetic field simulator for modeling the response of a radar sensor in an imaging system based on the synthetic aperture radar principle. The creation of completely random scenes is essential to achieve data sets with large variance. Therefore, special emphasis is placed on the development of methods that allow creating random objects, which can then be assembled into an entire scene. In the context of this contribution, we focus on humanitarian demining with regard to improvised explosive devices using a ground-penetrating radar system. This is an area where the use of trained classifiers is of great importance, but in practice, there are little to no labeled datasets for the training process. The simulation results show good agreement with the measurement results obtained in a previous contribution, demonstrating the possibility of enhancing sparse training data sets with synthetic data. https://lnkd.in/gQtMwmGN

9

Congratulations to the entire team at Noxon GmbH on successfully closing this seed funding round. I had the pleasure of meeting Maria Neugschwender about four years ago during the European Fascia Research Exchange Meeting (EFREM) at Technical University of Munich chaired by Prof. Dr. Robert Schleip and Werner Klingler. Even at that time, the technological vision behind Noxon GmbH stood out as a highly innovative and science-driven initiative. It is inspiring to see how this early vision has now evolved into a sophisticated platform capable of transforming human muscle activity into a digital interface for human–machine interaction. Such a technology holds tremendous promise for rehabilitation, neuromuscular disorders, and many other application domains. I would also like to congratulate HTGF | High-Tech Gründerfonds and Bayern Kapital for their continued commitment to identifying and supporting truly innovative deep-tech initiatives that shape the technologies of the future. As the founder of HGO MEDİKAL SAN. TİC. A.Ş., I sincerely hope that one day we may also see projects such as our LarvaSON (www.larvason.com) technology among the remarkable innovations supported by HTGF | High-Tech Gründerfonds. (LarvaSON is an eco-friendly ultrasonic platform designed to control mosquito larvae through targeted acoustic waves, replacing traditional chemical-based mosquito control methods.) Wishing the Noxon GmbH continued success on this exciting journey.

4

2 Kommentare

30 Years – 30 Stories! Foundation Models are transforming science. At Helmholtz, we’re developing powerful AI models that can analyze vast datasets and adapt to emerging challenges in health, energy, and the environment. Read more 👉 https://lnkd.in/ey8NZzUY Curious what Helmholtz Munich is working on in health research? Check out our Foundation Model story on how foundation models are shaping biomedical research: 👉 https://lnkd.in/djFZQSpx #30Years30Stories

10

Ok, okay. So, that Wednesday 10.12 - I participated in AI Tinkerers Bremen edition (always writing AI THINKERERS, god bless me). With pretty interesting topic. What if we take AxWise open source https://lnkd.in/eiGezPkf repository and generate synthetic dataset from WW2 with following entry criteria: "business_context": {   "business_idea": "Officer Recruitment & Defection Intelligence Network - Synthetic intelligence profiles of Wehrmacht and SS officers in occupied France/Germany (1943-1945) for Allied intelligence operations planning surrender negotiations, sabotage team infiltration, and identifying high-value defection targets. Generate realistic officer personas with personality profiles, command structure relationships, loyalty assessments, and negotiation vulnerabilities.",   "target_customer": "Allied Intelligence Operations: Intelligence Officer, Counterintelligence Operative, Sabotage Mission Planner Challenge: Identify which Nazi officers can be turned, which will resist to death, which hold strategic command",   "problem": "Cannot conduct traditional intelligence interviews with live Nazi officers (operational security, immediate execution risk). Need synthetic personality-consistent officer profiles to: (1) predict officer behavior in surrender scenarios, (2) identify defection leverage points, (3) plan infiltration team psychology, (4) anticipate officer objections in negotiation, (5) assess unit cohesion for sabotage vulnerability.",   "industry": "World War II Military Intelligence & Occupation Operations (1943-1945) Wehrmacht command structure across France, Austria, southern Germany SS/Gestapo organizational networks Military vs. Nazi party loyalties",   "location": "France: Paris, Lyon, Marseille (occupation zones, resistance networks) Austria: Vienna, Salzburg (Alpine redoubt planning) Germany: Berlin, Munich, Stuttgart (command centers, officer corps concentration) Italian Front: Post-armistice coordination zones" Imagine that we move there as time travelers in order to help Allies recruit some people from troublesome period of Germany with system that will help identify people who can make life of one austrian painter a bit harder. For sure that wasnt main point, but how accurate portrait of people who interested in that -will be for US OSS (pre-CIA org) members, like objections, who they are, what will be their pain points, challenges, how do they literally look like? What exactly AxWise solves - it targets users with whom you cannot talk or it is really hard to talk. Long story short, it was entertaining. People who understood it was about datasets - laughed, who didn't were a bit shocked because you know - it was kind of forced Godwins law. However, objections and mapping of US High Tier officiers from WW2 towards the system doesn't literally differ a lot from nowadays objections from modern Enterprise (what an irony). In the end it is just a context engineering. Thanks a lot JAAI | JUST ADD AI GmbH

14

2 Kommentare

New DFG-funded project at the Chair of Automated and Autonomous Systems (AAS), in cooperation with the Deutsche Sporthochschule Köln! Our interdisciplinary project CoExo, funded by the Deutsche Forschungsgemeinschaft (DFG) - German Research Foundation (https://lnkd.in/ea7FPAuF), aims to investigate and optimise intelligent exoskeleton assistance to counteract cognitive-motor interference (CMI) in real-world industrial applications. Exoskeletons are innovative ergonomic tools designed to support workers by supporting movements or stabilising postures. However, CMI—the competition between cognitive and physical tasks for limited neural resources—poses a significant challenge affecting overall worker performance. At AAS, we will develop a range of context-dependent intelligent levels — from passive systems to adaptive AI-driven support functions defined in successive steps — that are embedded in our self-developed prototype exoskeleton “Lucy”. Our project partner, the Chair of the Institute of Movement Therapy and Movement-Oriented Prevention and Rehabilitation (https://lnkd.in/e2wV2p_N), led by Prof. Dr. Bettina Wollesen, will conduct comprehensive experimental evaluations, employing advanced techniques such as 3D motion capture, electromyography (EMG), and functional near-infrared spectroscopy (fNIRS) to analyse cognitive and motor responses. For further information, please see the attached project profile! Robert Weidner Bettina Wollesen David Scherb Amir Nemati Daniel Büchel Franziska Schmitter TU Bergakademie Freiberg Deutsche Sporthochschule Köln Deutsche Forschungsgemeinschaft (DFG) - German Research Foundation

31

Dear Jensen Huang, at the most recent “Tag der Industrie 2025” with chancellor Friedrich Merz, you claimed that we Germans are unable to develop software because we are too “precise.” 𝗜 𝗱𝗶𝘀𝗮𝗴𝗿𝗲𝗲: 𝗽𝗿𝗲𝗰𝗶𝘀𝗶𝗼𝗻 𝗶𝘀𝗻’𝘁 𝗮 𝘄𝗲𝗮𝗸𝗻𝗲𝘀𝘀. 𝗜𝗻 𝗼𝘂𝗿 𝗰𝗮𝘀𝗲, 𝗶𝘁’𝘀 𝗮 𝗿𝗲𝗾𝘂𝗶𝗿𝗲𝗺𝗲𝗻𝘁 𝗳𝗼𝗿 𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗮𝗻𝗱 𝗱𝗮𝘁𝗮 𝘁𝗵𝗮𝘁 𝗹𝗮𝘀𝘁𝘀. I got your point, Jensen, but please take into account: Someone who works accurately can always botch things up at any time. A bungler, on the other hand, can never work accurately. “Doing the right thing right” is a challenge and a vision, but it is the only possible position - it is the only sustainable way in our branch out of digital chaos. We’re even willing to sacrifice our short-term gains (profits) to get there. We see it every day: if wrong software is used or software isn’t built with care, users often end up with messy, inconsistent, unstructured data. That’s not value - it’s data waste. 𝗧𝗵𝗶𝘀 𝗸𝗶𝗻𝗱 𝗼𝗳 𝗱𝗮𝘁𝗮 𝗰𝗲𝗿𝘁𝗮𝗶𝗻𝗹𝘆 𝗶𝘀 𝗻𝗼𝘁 “𝘁𝗵𝗲 𝗻𝗲𝘄 𝗼𝗶𝗹”. Only 𝘷𝘢𝘭𝘪𝘥 data is. And so this might create the demand for more processing power, but it’s hardly progress. Real progress means creating clean, usable data right from the start. But don’t worry - we are using NVIDIA chips in a much smarter way than just fixing “data chaos” with the help of AI. At primtech, we develop 𝗱𝗮𝘁𝗮𝗯𝗮𝘀𝗲-𝗱𝗿𝗶𝘃𝗲𝗻 𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗳𝗼𝗿 𝗰𝗿𝗶𝘁𝗶𝗰𝗮𝗹 𝗶𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲. There’s no room for “move fast and break things” when the cost of bad data is real-world damage. In other industries, this might work. In many use cases, it would be fatal. We use AI to build valid Digital Twins of existing HV substations. Of course, even our German employees are able to deal with chaos and faulty data. But if we have a chance, we do it right from the start - with precision, and care, and valuable data. Our customers don’t need more silos, they need clarity and consistency. “German thoroughness” isn’t a bottleneck. It’s a strength. Especially in the energy sector, where many global engineering and software leaders come from Germany - built on deep expertise, long-term thinking, and quality by design. That’s also how we work at entegra gmbh. People from over 17 nations contribute their skills and knowledge to build software that delivers on its promise. Immigrant expertise is part of our success. We employ industry experts - not just to build the product, but to support it. People who understand the project, the industry, and the real challenges behind the screens. If you think precision slows things down, you might be misunderstanding what digitalization is actually about. PS: Funny enough, you also advised young people to focus on physical sciences instead of software. Seems like you know that real-world impact still relies on precision and expertise. Welcome back to reality. #Innovation #DigitalTransformation #AI #BusinessModelInnovation #Germany

41

5 Kommentare

🏆 Last week we had the final presentation of project GreenEdge-FuE, where we investigated methods to execute AI models for object detection directly on traffic intersections, instead of streaming the images to some centralized server and processing them there.

16

2 Kommentare

European robotics is no longer concentrated. It’s fragmenting. But that’s a good thing. A few years ago, most “EU robotics” conversations pointed at one or two hubs. Now it’s regional, with clusters getting deeper: 🔹 Zurich and the Swiss cluster ANYbotics ABB ETH Zürich 🔹 Southern Germany cluster KUKA STÄUBLI Bosch Rexroth Festo Baumer Group 🔹 Stuttgart region cluster NEURA Robotics Bosch TRUMPF 🔹 Munich and Bavaria cluster Agile Robots SE Franka Robotics 🔹 Odense (Denmark) cluster Odense Robotics Universal Robots Mobile Industrial Robots 🔹 Lille / Paris (France) cluster Exotec Wandercraft Naïo Technologies Fragmentation matters because it changes the economics of shipping robots. When customers are closer, you iterate faster. When field support is local, uptime becomes normal. When integrators and suppliers sit in the same region, deployments stop resetting every time you cross a border. The next winners won’t be the loudest, they’ll be the densest. Clusters win because support and integration are local. It’s impossible to tag everyone in one post, so feel free to add the companies I missed that are building serious robots in Europe. #Robotics #Automation #IndustrialAutomation #DeepTech #EuropeanTech #AdvancedManufacturing Stuart Cooper

64

22 Kommentare

Excited to share what we at AI Marketplace (KI-Marktplatz) have been working on as part of the #FAICE project by prostep ivip Association 🚀 Our team developed a #SysML Assistant prototype that shows how AI can transform Model-Based Systems Engineering (MBSE) – making it faster, more reliable, and easier to adopt. This is just one example of how we’re turning AI research into practical engineering solutions. At AI Marketplace, we’re exploring many such innovations that make engineering #smarter and more #collaborative. Curious about what else we’re building? DM us or Check out more on our website: https://ki-marktplatz.com/ Ruslan Bernijazov | Rik Rasor | Roman Dumitrescu | Namit Joshi | Aryman Deshwal | Mansoor Soomro #AI #Engineering #SysML #MBSE #AIinEngineering #KIMarktplatz #AIMarketplace #InnovationInGermany

29

2 Kommentare

"The US Cloud Act allows American authorities to access data stored by US companies – even if it is stored on servers outside the USA. For German research, this means legal uncertainty, a lack of control and, in an emergency, the loss of central bodies of knowledge."

15

2 Kommentare

Our (Bundesdruckerei GmbH) research papers together with Multiverse Computing are out on the arXive. 1- Quantum Inspired, prrivacy aware synthetic data generation Using Matrix Product States (MPS) we generate realistic, privacy preserving structured data with a new state-of-the-art model. https://lnkd.in/e-ttn8r7 2- Blockchain Network Analysis with quantum inspired GNNs and QBoosting We developed a framework based on GNNs and quantum inspired/qboosting techniques to detect anomalous entities and flows in Blockchain networks. https://lnkd.in/ef7cRD6z I am very excited to see how methods like tensor networks that were originally designed in physics can bring advantages in scalability, interpretability, and efficiency of ML models that are specifically relevant for data privacy and anomaly detection. These works were the outcome of collaboration between our team from the Innovation Hub of Bundesdruckerei-Gruppe and Multiverse Computing. Special thanks to Holger Eble, Manuel Radons, Josef Menter, Oliver Muth, and Kim Nguyen; and the MVC team: Alejandro Moreno Rodríguez, Daniel De Rosso, Desale Fentaw Nurye, Luigi D'Amico, Ninad Dixit, Samuel Palmer, Raul Salles de Padua, Samuel Mugel, PhD, Román Orús , and Victor Gaspar Curious to hear more about our works on quantum and quantum inspired AI works, please come to our presentations and our joint workshop with MVC at Bitkom Events - quantum summit!

54

4 Kommentare

Spec-driven development with coding agents: does it kill agile methods and bring us back to waterfall? That’s the question Wolfgang Strunk left us with at the end of his guest lecture today about agile architecture, which I had the pleasure of hosting. I’m closely watching and actively participating in the changes coding agents are bringing to our daily work. They’re powerful tools with real potential to boost productivity, but they also come with challenges that teams will need to navigate carefully. So, will agile methods survive? I believe they will. Iterative processes remain essential since we still need to fully understand what customers and users actually want (as opposed to what they think they want). That said, the way we practice agile will likely evolve. It will be fascinating to see how teams navigate this transition, integrate new tools, and what it means for software engineering in practice. What do you think will change? My guess: documentation will rise in value as it sets the context for coding agents. #Agile #SpecDrivenDevelopment #CodingAgents #SoftwareEngineering #FutureOfWork #AIinSoftware

9

5 Kommentare

⚛️ It's #Physics Time: Sommerfeld Expansion - Unlocking Fermions at Low Temperatures ⚛️ 📜 A Glimpse Back in Time In 1928, Arnold Sommerfeld extended Fermi’s ideas about quantum statistics to better describe the behavior of electrons in metals. His key tool, now known as the Sommerfeld expansion, provided physicists with a way to handle complex integrals involving the Fermi-Dirac distribution at very low temperatures. This breakthrough allowed for precise predictions of heat capacities, conductivities and many other electronic properties in condensed matter physics. 🧩 What It’s All About At the heart of the problem lies the Fermi-Dirac distribution, which governs how fermions (like electrons) occupy energy states. At absolute zero, all states up to the Fermi energy are filled and all above are empty (so it's a Heavyside function). But at low (but finite) temperatures, this step function “smears out.” Directly calculating physical quantities becomes very challenging, because integrals over the distribution are difficult to solve. The Sommerfeld expansion provides an elegant approximation method: it expands the integrals in powers of temperature, capturing the leading corrections beyond the zero-temperature limit. 📸 In the attached photo, you can see the derivation worked out step by step - from the Fermi-Dirac distribution, through Taylor expansions and special functions, to the final compact form of the expansion. ⚙️ Where It Shows Up The Sommerfeld expansion isn’t just a mathematical trick, it’s a powerful tool with wide-ranging applications: 1️⃣ Electron heat capacity: Explains why the electronic contribution to the heat capacity of metals is linear in temperature. 2️⃣ Electrical conductivity: Connects transport properties with the density of states near the Fermi level. 3️⃣ Magnetic susceptibility: Describes how conduction electrons respond to magnetic fields. 4️⃣ Quantum gases: Provides approximations for trapped fermionic atoms in ultracold experiments. 🚀 Looking Ahead The Sommerfeld expansion remains foundational, but new frontiers - from strongly correlated electrons to exotic fermionic matter in astrophysics - push beyond its reach. Understanding these limits and developing generalizations continues to be an active area of research. ✅ In a Nutshell The Sommerfeld expansion is one of those “simple but powerful” methods: a cornerstone of theoretical physics that helps bridge abstract quantum statistics and real-world material properties. #Physics #CondensedMatter #QuantumMechanics #FermiDirac #SommerfeldExpansion #LowTemperaturePhysics

64

3 Kommentare

News from Germany’s innovation agency SPRIND (https://www.sprind.org/) : After two‑and‑a‑half years of prototyping, the Long‑Duration Energy Storage Challenge ends with four winners. Who made the cut? Reverion - hydrogen‑based solid‑oxide system. Ore Energy - iron‑air battery. HalioGen Power - membrane‑less flow battery. Unbound Potential - membrane‑less flow battery (different architecture). ⚡ Why it matters Long‑duration storage (10–100 h) is the missing puzzle piece of a 100 % renewable grid. While Berlin debates building 40–50 new gas plants to ride through dark wind‑less weeks, SPRIND showcased alternative solutions. 👉 Which tech do you think will scale first? Source: Rico Grimms Cleantech Newsletter #EnergyStorage #CleanTech #Hydrogen #IronAir #FlowBatteries #ClimateTech #Dunkelflaute

3

About how deep tech makes its way into industry; while spending time around the research ecosystem in Germany, I noticed something interesting. I saw moments where research, applied institutes like Deutsches Forschungszentrum für Künstliche Intelligenz (DFKI), and industry were at least in the same room. The handoff wasn’t seamless or guaranteed, but there was an observable effort to bridge research and real-world use. I felt the same dynamic during SSDA ’23 at Université de Fribourg - Universität Freiburg, where researchers and industry practitioners were actively exchanging constraints. Not everyone gets access to that kind of proximity. And even when they do, adoption is still hard especially in slow-moving industries. That’s where many deep-tech startups struggle today. What the deep-tech startups that collaborate with us during the build & production phase tend to get right is this: they don’t wait for industry validation. They manufacture it by - running short pilots. - bench-marking against existing industrial systems. - measuring things industry actually cares about: latency, power, reliability, integration. Those comparisons change the conversation. That’s often the difference between pure research and real adoption. If you’re building in robotics, neuromorphic systems, or embedded AI, what’s the hardest thing for you to prove before industry will even engage?

56

1 Kommentar