About
Activity
10K followers
-
Arno Candel shared thisSuper proud to have worked on that project, first during my PhD thesis at ETH Zürich and then as a staff scientist at SLAC National Accelerator Laboratory! #Supercomputing Thanks to my PhD advisers Ralph Eichler and Matthias Troyer for guidance and foresight which brought me to the US. All that before AI/ML took over my world :) Even better, FELs may end up powering how we make future AI chips!Arno Candel shared thisA proud part of decades of #DOE innovation As one of the U.S. Department of Energy (DOE) National Labs, SLAC is home to the world’s first X-ray Free-Electron Laser, our Linac Coherent Light Source (LCLS). Its snapshots of atoms and molecules at work help us advance technologies of the future, develop life-saving drugs & innovate energy solutions. We’re celebrating #Freedom250 and beyond by honoring our history and preparing for what comes next.How we pioneered the first X-ray free-electron laserHow we pioneered the first X-ray free-electron laser
-
Arno Candel reposted thisArno Candel reposted thisGoldilocks Program and some Miracles (5) Is this equation worth $4.7 trillion? That's the current market value of NVIDIA. But NVIDIA GPUs may soon be obsolete. TL;DR: Deep learning without training. No randomness. No gradients. No GPUs. No matrices. Learning = single analytic geometric projection. Imagine training a model in single analytic step, without randomness, on your laptop, everytime reproducible, tuned to preserve explicit geometric invariants, and an accuracy of choice anywhere up to 100%. This is today's Goldilock's "miracle" - a result enabled by the Riemann zeta function's log prime basis (Euler prime product). Hence, you don't want to miss this new paper: “Deterministic Deep Representation Learning via Geometric Invariant Tuning” https://lnkd.in/gnpsnMwZ ➡️ One-step projection: analytic tangent-space solution - no iterative training, no loss landscape wandering. ➡️ Matrix-free & embarrassingly parallel: complexity O(N log N), not O(N³). ➡️ Reproducible by design: same data leads to same representation, always. ➡️ Geometric invariants: moments, centroids, and zero-crossings preserved exactly. ➡️ Energy-efficient: many orders of magnitude less energy than CNNs. ➡️ Accuracy-dialing: one parameter smoothly trades compression for accuracy (“the compression cliff”). This podcast breaks it down too: https://lnkd.in/gwqqbZuc Happy AI training! P.S.: I'm still recruting exceptional minds to join the Goldilocks Research Program. DM me for more details. #AI #MachineLearning #DeepLearning #ArtificialIntelligence #DataScience #Reproducibility #EdgeAI #EnergyEfficiency #Innovation #Research Octonion Group Michael Erlihson, PhD Colleen Farrelly Srini Pagidyala Dimitrios A. Karras Derrick Hodge Patrick Nicolas
-
Arno Candel shared thisProud to have worked on one of the most iconic inventions from the past 25 years! #CERN #SLAC #ETHZ #PSI https://lnkd.in/gzQshdceLHC inducted into TIME’s Best Inventions Hall of FameLHC inducted into TIME’s Best Inventions Hall of Fame
-
Arno Candel shared thisGot cited by a Nature paper (for H2O-3 OSS machine learning https://lnkd.in/gXR6KJ8H)! https://lnkd.in/g7VztjUKA specific gene expression program underlies antigen archiving by lymphatic endothelial cells in mammalian lymph nodes - Nature CommunicationsA specific gene expression program underlies antigen archiving by lymphatic endothelial cells in mammalian lymph nodes - Nature Communications
-
Arno Candel shared thisH1-B is how I started my first job as a Staff Scientist at Stanford's Linear Accelerator 20 years ago, right after graduating from ETHZ with a PhD in physics and supercomputing. Three years ago, I became a US citizen to complete the cycle. Thanks to everyone who made this possible, you are not forgotten.
-
Arno Candel shared thisYou can buy this Kaggle-Grandmaster-in-a-box AutoML platform already, it's called H2O Driverless AI. In the cloud, or on-premises, GPU accelerated. Saving banks and telcos billions a year already. We spent 5 years at H2O.ai to build it, with 2x NVIDIA funding. One of the best kept secrets in Enterprise AI IMO.Arno Candel shared thisAfter hundreds of Kaggle competitions and years of trial, error, and wins, our team Kazuki Onodera, Théo Viel, Chris Deotte and I, have compiled a playbook for tabular modeling—a system that’s carried us to the top of leaderboards and held up on real-world data🥳 . We’ve distilled what we learned into 7 battle-tested techniques: - Smarter EDA: spot shifts and patterns most people miss. - Diverse baselines: compare models early to see the landscape. - Feature engineering at scale: thousands of features, not dozens. - Ensembling: Hill climbing + Stacking to combine model strengths. - Pseudo-labeling: turn unlabeled data into training signal. - Extra training: multiple seeds + full-data retraining for the final gains. All of these rely on fast experimentation (made practical with GPUs) and careful validation. Here’s the full playbook with links to solution write-ups and notebooks: https://lnkd.in/dH-Xr3Nw This is years of our team’s work, experiments, and lessons learned—all in one place. Hopefully it helps you climb your own leaderboard 🚀 , whether that’s on Kaggle or in production. #nvidia #machinelearning #kaggle #rapidsaiThe Kaggle Grandmasters Playbook: 7 Battle-Tested Modeling Techniques for Tabular Data | NVIDIA Technical BlogThe Kaggle Grandmasters Playbook: 7 Battle-Tested Modeling Techniques for Tabular Data | NVIDIA Technical Blog
-
Arno Candel shared this+1Arno Candel shared thisThere is significant unmet demand for developers who understand AI. At the same time, because most universities have not yet adapted their curricula to the new reality of programming jobs being much more productive with AI tools, there is also an uptick in unemployment of recent CS graduates. When I interview AI engineers — people skilled at building AI applications — I look for people who can: - Use AI assistance to rapidly engineer software systems - Use AI building blocks like prompting, RAG, evals, agentic workflows, and machine learning to build applications - Prototype and iterate rapidly Someone with these skills can get a massively greater amount done than someone who writes code the way we did in 2022, before the advent of Generative AI. I talk to large businesses every week that would love to hire hundreds or more people with these skills, as well as startups that have great ideas but not enough engineers to build them. As more businesses adopt AI, I expect this talent shortage only to grow! At the same time, recent CS graduates face an increased unemployment rate, though the underemployment rate — of graduates doing work that doesn’t require a degree — is still lower than for most other majors. This is why we hear simultaneously anecdotes of unemployed CS graduates and also of rising salaries for in-demand AI engineers. When programming evolved from punchcards to keyboard and terminal, employers continued to hire punchcard programmers for a while. But eventually, all developers had to switch to the new way of coding. AI engineering is similarly creating a huge wave of change. There is a stereotype of “AI Native” fresh college graduates who outperform experienced developers. There is some truth to this. Multiple times, I have hired, for full-stack software engineering, a new grad who really knows AI over an experienced developer who still works 2022-style. But the best developers I know aren’t recent graduates (no offense to the fresh grads!). They are experienced developers who have been on top of changes in AI. The most productive programmers today deeply understand computers, how to architect software, and how to make complex tradeoffs — and who additionally are familiar with cutting-edge AI tools. Sure, some skills from 2022 are becoming obsolete. For example, a lot of coding syntax that we had to memorize back then is no longer important, since we no longer need to code by hand as much. But even if, say, 30% of CS knowledge is obsolete, the remaining 70% — complemented with modern AI knowledge — is what makes really productive developers. Without understanding how computers work, you can’t just “vibe code” your way to greatness. Fundamentals are still important, and for those who additionally understand AI, job opportunities are numerous! [Original text: https://lnkd.in/gHSWcKhr ]
-
Arno Candel shared thisInteresting that OpenAI calls their open models 'gpt-oss' and 'open-weight' simultaneously. They are contributing to the open source software (oss) ecosystem, but the models are not open source, as there's no data or training code. https://lnkd.in/gWFH8zXY
-
Arno Candel liked thisArno Candel liked thisAn unforgettable day at the NYSE. 🔔 So special to be there to watch Andrew Dudum ring the Opening Bell with his wife and kids alongside the Hims & Hers team. Proud to be part of this incredible team and milestone - excited for what’s ahead!
-
Arno Candel liked thisArno Candel liked this📣 Today we're launching H2O-3 Secure. For more than a decade, H2O-3 has been the open source AI/ML foundation for 2M data scientists, Kagglers, and ML practitioners worldwide. That community is the reason H2O.ai exists. What we are now seeing is that the same data scientists who built on H2O-3 are being asked to prove production controls: vulnerability management, incident response, and vendor risk documentation. They need a solution that is geared for production deployment and provides assurance around scalability, auditability (SOC 2, ISO 27001, and ISO 42001) and is supported 24x7 by a team of experts. Our answer is H2O-3 Secure 🔒, API compatible with the open source version, that provides these assurances as well as CVE patching and long-term support. H2O-3 Secure is built on the same security controls as H2O.ai Cloud for Government (H2O-GOV), extending FedRAMP-grade rigor to the H2O-3 workloads users run in production. 🛠️ A word to the H2O-3 OSS community: H2O-3 OSS remains free and Apache 2.0-licensed. If you need production MOJO model scoring for Hadoop, Spark, Teradata or Oracle, we can provide a free license for non-commercial use. Request it at https://h2o.ai/h2o-3/ What changes: we've written up exactly what this means for OSS users here: https://lnkd.in/gd8zV_bE Schedule an H2O-3 Production Review to see which path is right for you: h2o.ai/h2o-3/oss-vs-secure #H2Oai #H2O3 #OpenSource #AI
-
Arno Candel reacted on thisArno Candel reacted on thisWatching Apollo 16 astronaut Charlie Duke —a man who walked on the Moon and a true friend of my wife, Sherry Epperson —present her with the Air Show League Founding Fellow Award was incredibly moving. Even more remarkable was knowing that Sherry was the very first person to ever receive this honor. Nearly thirty years ago, she agreed to help a friend with what was supposed to be a nine-month job. She never imagined it would become her life’s work. Since then, she has helped shape America’s presence at the world’s premier air shows, earning the respect of industry leaders, military commanders, U.S. embassies, and international partners through integrity, hard work, and quiet leadership. She has represented American aerospace at air shows across Paris, Farnborough, Dubai, Singapore, Avalon, and beyond. Together with fellow Founding Fellow Mike Olson, Sherry helped create what became the International Rocketry Challenge, bringing STEM students from five countries together through science, engineering, and friendly competition. That same passion continued through the American Rocketry Challenge, the Propel Science Foundation Science Foundation, United States Naval Academy Sponsor family, where she has mentored hundreds of young people. Many have gone on to become engineers, commercial and military aviators, military officers, and industry leaders. She has always believed our greatest legacy isn’t what we accomplish ourselves, but the people we inspire. Those who know Sherry know she has never chased titles or recognition. She simply shows up, works tirelessly, lifts others up, and makes everyone around her better. Her career has been defined by building international partnerships, promoting American aerospace, and creating opportunities for others to succeed. Today, that lifetime of service was recognized by the people who know this industry best—the CEOs of the world’s leading international air shows, together with Kallman Worldwide, Inc. . There is no greater compliment than the respect of your peers. Being the very first recipient of the Air Show League Founding Fellow Award is extraordinary, but it isn’t what makes me proudest. I’m proudest of the friendships you’ve built, the students you’ve inspired, and the example you’ve set for everyone fortunate enough to know you. As I watched Charlie present you with this honor, I couldn’t help but think of your dad. He always believed in you and encouraged you to dream big. I know he would have been incredibly proud to see his daughter recognized by the leaders of the global aerospace community and receive this award from one of history’s greatest explorers. A special thank you to Tom Kallman for creating the Air Show League Founding Fellow Award and recognizing those whose dedication has helped shape the international aerospace community. Your vision has created a lasting legacy. No one is more deserving. I love you.
-
Arno Candel liked thisArno Candel liked thisHuge news from our Nemo Retriever team at NVIDIA! 🚀 We've officially released a project we've been actively working on lately. Meet Nemotron-3-Embed, a new frontier family of open, commercial text embedding models designed for ultimate performance and efficiency. Here is the model lineup: 🏆 Nemotron-3-Embed-8B-BF16 is our flagship embedding model, which currently ranks #1 on both Public and Private splits of the RTEB Leaderboard 🧪 Nemotron-3-Embed-1B-BF16 is a pruned and distilled model built for efficiency, retaining 95%+ of the 8B teacher's accuracy making it SOTA in its parameter range 🏎️ Nemotron-3-Embed-1B-NVFP4 is the Blackwell-optimized variant for ultra-high-throughput retrieval while retaining 99.5% of BF16 accuracy 📖 Read the blog post with the details: https://lnkd.in/dWgyxTuK 🤗 Try the models on Hugging Face and leave a ❤️: https://lnkd.in/dV7u-Sqx
-
Arno Candel liked thisArno Candel liked this#SiliconValley part 2 It is not a location, it is a mindset. Bill Wright inspired us with a coffee mug ☕ , yes, a simple coffee mug, that made all the difference in a situation where he first thought "it is not gonna work, it is too big an idea"...he went for it anyway, and today he is the chair of his brainchild, Enterprise Neurosystem. Christie Ko, Executive Director of the Stanford Digital Economy Lab brought a vibrant energy to the classroom that perfectly reflected the Lab's spirit: a place she describes as one where "there's never a dull moment," driven by diverse disciplines and a shared excitement for building the future, with research spanning the economics of #AI, digital platforms, and new measures of prosperity. Stanford graduate and former #NFL player, Allen Smith proves that the discipline of the football field translates seamlessly into venture capital. As Founder and Managing Partner of Musa Capital, he backs diverse founding teams building transformative AI solutions, while providing underrepresented founders not just with capital, but with the operational support to reach the next level. With a PhD from MIT and over 25 years in industry, spanning Lawrence Berkeley Labs, IBM, and AT&T Bell Labs, Stanford Professor Jimmy Chen brought a wealth of real-world energy expertise to the classroom as Managing Director of the StorageX Initiative at Stanford's Precourt Institute for Energy, where he leads pioneering programs in hydrogen, energy storage, and decarbonization, and builds global communities connecting academia, industry, and government to accelerate the clean energy transition. What these lectures reminded us of, things we may have known or sensed before, but experiencing them in this setting made it real and shifted something: ✅ Say "f..k it" more often and just go for it ☕ ✅ We can't do "stuff" alone, connections are everything ✅ AI could make the economic pie much bigger. It has the potential to create enormous new wealth and opportunity, but it won't reach everyone automatically, staying curious and engaged is how you stay part of it ✅ The best talent is often the most overlooked, and that's exactly where the opportunity is ✅ Pioneer - Integrate - Propel The campus tour ended the day the best way possible. Reuniting with my childhood and teenage friend Arno Candel, who has lived and worked in Palo Alto for over 20 years, and diving into one of those rare, deep conversations about AI, the future, and whether there might just be a formula that can explain it all. Stanford University Stanford GSB Executive Education Executive MBA University of Zurich Tamara Carleton Thorsten Klein Carol Fuzzard STANFORD FACULTY CLUB
-
Arno Candel liked thisArno Candel liked thisI’m excited to share that I’ve joined Lambda as Account Director, Superintelligence. I’m grateful for my time at VAST Data and for the opportunity to work with strong teammates, customers, and partners across the AI infrastructure ecosystem. A big thank you to Robert Brooks IV and Eugene Berson for the opportunity and for the thoughtful conversations throughout the process. I’m looking forward to working closely with Sam Khosroshahi and the broader Lambda team as we help strategic AI customers scale infrastructure for some of the most demanding workloads in the market. It’s an exciting time to be working at the intersection of AI, compute, and customer strategy — where performance, reliability, and efficiency all matter. At Lambda, I’m excited to support customers who are pushing the frontier of AI and looking to maximize tokens per dollar per watt as they scale. Excited for what’s ahead!
-
Arno Candel liked thisArno Candel liked thisToday my youngest son taught me a simple lesson. We were heading into Jersey Mike’s for lunch when we saw a man sitting outside with a paper cup and a sign that read, “I’ll tell you 5 jokes.” He wasn’t asking anyone for money but it was clear he would benefit from some help. As we walked in, my son looked at me and said, “I wonder if he’s hungry.” That simple observation resonated with me, and I was thinking the same thing, so I ordered an extra sandwich. When we came back out, I asked the man if he was hungry and offered it to him. His eyes lit up. “Thanks, you rock, man!”, he said. Then he delivered on his promise: “What do you call a pig that’s learning karate?” ………….. “Pork Chop!” LOL! My son and I were laughing all the way to the car. The whole interaction lasted maybe a minute, but everyone walked away a little happier. As parents, we spend a lot of time trying to teach our kids values like empathy, kindness, and generosity. Sometimes, though, they’re the ones teaching us. Kids often notice the human need before we do. A small act of kindness cost very little, but it made a difference—and not just for the person receiving it. Keep spreading kindness. Listen to your kids. And never pass up a good joke! #Kindness #Leadership #Parenting #Empathy
-
Arno Candel liked thisIt’s amazing to see how we can accelerate R&D with agentic AI and see the amazing improvements with the Majorana 2 design.Arno Candel liked thisIt has been a fantastic morning at Microsoft with Satya Nadella at Build. 🙌 Introducing Majorana 2: I’m incredibly proud of what this team has delivered in record time - a next‑generation quantum chip that marks a true shift from physics to engineering. Majorana 1 proved the physics; Majorana 2 proves the engineering, with a level of reliability the field has never seen. What makes this achievement even more extraordinary is how deeply agentic AI accelerated the work - from materials discovery to device optimization - enabling breakthroughs that would have taken years by traditional methods. This is Microsoft Discovery-enabled, built through partnership, and I want to express my appreciation for Jason Zander’s leadership and supporting our roadmap. Moments like this remind me how rare it is to work with people who can push the boundaries of science and engineering at the same time and do it with this level of rigor and ambition. Thank you Matthias Troyer, Chetan Nayak, Lauri Sainiemi, and everyone at Microsoft Quantum for everything you’ve poured into this. Read more here 👉 https://lnkd.in/gKRhJnVy
Experience & Education
Volunteer Experience
-
Teaching Assistant
Stanford Continuing Studies
- 2 years 4 months
Arts and Culture
Assisted photography classes with Joel Simon http://www.joelsimonimages.com
Publications
-
NUMERICAL VERIFICATION OF THE POWER TRANSFER AND WAKEFIELD COUPLING IN THE CLIC TWO-BEAM ACCELERATOR
SLAC
See publicationThe Compact Linear Collider (CLIC) provides a path to a multi-TeV accelerator to explore the energy frontier of High Energy Physics. Its two-beam accelerator (TBA) concept envisions complex 3D structures, which must be modeled to high accuracy so that simulation results can be directly used to prepare CAD drawings for machining. The required simulations include not only the fundamental mode properties of the accelerating structures but also the Power Extraction and Transfer Structure (PETS), as…
The Compact Linear Collider (CLIC) provides a path to a multi-TeV accelerator to explore the energy frontier of High Energy Physics. Its two-beam accelerator (TBA) concept envisions complex 3D structures, which must be modeled to high accuracy so that simulation results can be directly used to prepare CAD drawings for machining. The required simulations include not only the fundamental mode properties of the accelerating structures but also the Power Extraction and Transfer Structure (PETS), as well as the coupling between the two systems. Time-domain simulations will be performed to understand pulse forma- tion, wakefield damping, fundamental power transfer and wakefield coupling in these structures. Applying SLAC’s parallel finite element code suite, these large-scale prob- lems will be solved on some of the largest supercomputers available. The results will help to identify potential issues and provide new insights on the design, leading to further improvements on the novel two-beam accelerator scheme.
-
State of the art in electromagnetic modeling for the Compact Linear Collider
SLAC
See publicationSLAC’s Advanced Computations Department (ACD) has developed the parallel 3D electromagnetic time-domain code T3P for simulations of wakefields and transients in complex accelerator structures. T3P is based on state-of-the-art Finite Element methods on unstructured grids and features unconditional stability, quadratic surface approximation and up to 6th-order vector basis functions for unprecedented simulation accuracy. Optimized for large-scale parallel processing on leadership supercomputing…
SLAC’s Advanced Computations Department (ACD) has developed the parallel 3D electromagnetic time-domain code T3P for simulations of wakefields and transients in complex accelerator structures. T3P is based on state-of-the-art Finite Element methods on unstructured grids and features unconditional stability, quadratic surface approximation and up to 6th-order vector basis functions for unprecedented simulation accuracy. Optimized for large-scale parallel processing on leadership supercomputing facilities, T3P allows simulations of realistic 3D structures with fast turn-around times, aiding the design of the next generation of accelerator facilities. Applications include simulations of the proposed two-beam accelerator structures for the Compact Linear Collider (CLIC) – wakefield damping in the Power Extraction and Transfer Structure (PETS) and power transfer to the main beam accelerating structures are investigated.
-
HIGH-FIDELITY INJECTOR MODELING WITH PARALLEL FINITE ELEMENT 3D ELECTROMAGNETIC PIC CODE PIC3P
SLAC
See publicationSLAC’s Advanced Computations Department (ACD) has developed the parallel Finite Element 3D electromagnetic code suite ACE3P for modeling of complex accelerator structures. The Particle-In-Cell module Pic3P was designed for simulations of beam-cavity interactions dom- inated by space charge effects. Pic3P solves the complete set of Maxwell-Lorentz equations self-consistently and in- cludes space-charge, retardation and boundary effects from first principles. In addition to using conformal…
SLAC’s Advanced Computations Department (ACD) has developed the parallel Finite Element 3D electromagnetic code suite ACE3P for modeling of complex accelerator structures. The Particle-In-Cell module Pic3P was designed for simulations of beam-cavity interactions dom- inated by space charge effects. Pic3P solves the complete set of Maxwell-Lorentz equations self-consistently and in- cludes space-charge, retardation and boundary effects from first principles. In addition to using conformal, unstructured meshes in combination with higher-order Finite Element methods, Pic3P also uses causal moving window techniques and dynamic load balancing for highly efficient use of computational resources. Operating on work-stations and on leadership-class supercomputing facilities, Pic3P allows large-scale modeling of photoinjectors with unprecedented accuracy, aiding the design and operation of next-generation accelerator facilities. Applications include the LCLS RF gun.
Honors & Awards
-
Featured in Swiss National Radio
SRF1
https://t.co/wUTUXxsqU7
-
Featured in Wired Magazine
Wired.com
https://www.wired.com/2017/01/move-coders-physicists-will-soon-rule-silicon-valley/
-
Featured in ETH Globe Q3/2015
ETH Zurich
https://www.ethz.ch/content/dam/ethz/common/docs/media/globe/2015/eth-globe-3-2015-en.pdf
-
Fortune Magazine 2014 Big Data All Star
Fortune Magazine
http://fortune.com/2014/08/03/meet-fortunes-2014-big-data-all-stars/
-
Personal Letter from Governor Arnold Schwarzenegger
State Capitol, Sacramento, California
-
Summa Cum Laude
ETH Zurich
Languages
-
English
Full professional proficiency
-
German
Native or bilingual proficiency
-
French
Limited working proficiency
-
Turkish
Elementary proficiency
-
Swiss German
Native or bilingual proficiency
-
Dutch
Elementary proficiency
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content