Engineering

Exciting updates on Project GR00T! We discover a systematic way to scale up robot data, tackling the most painful pain point in robotics. The idea is simple: human collects demonstration on a real robot, and we multiply that data 1000x or more in simulation. Let’s break it down: 1. We use Apple Vision Pro (yes!!) to give the human operator first person control of the humanoid. Vision Pro parses human hand pose and retargets the motion to the robot hand, all in real time. From the human’s point of view, they are immersed in another body like the Avatar. Teleoperation is slow and time-consuming, but we can afford to collect a small amount of data.  2. We use RoboCasa, a generative simulation framework, to multiply the demonstration data by varying the visual appearance and layout of the environment. In Jensen’s keynote video below, the humanoid is now placing the cup in hundreds of kitchens with a huge diversity of textures, furniture, and object placement. We only have 1 physical kitchen at the GEAR Lab in NVIDIA HQ, but we can conjure up infinite ones in simulation. 3. Finally, we apply MimicGen, a technique to multiply the above data even more by varying the *motion* of the robot. MimicGen generates vast number of new action trajectories based on the original human data, and filters out failed ones (e.g. those that drop the cup) to form a much larger dataset. To sum up, given 1 human trajectory with Vision Pro  -> RoboCasa produces N (varying visuals)  -> MimicGen further augments to NxM (varying motions). This is the way to trade compute for expensive human data by GPU-accelerated simulation. A while ago, I mentioned that teleoperation is fundamentally not scalable, because we are always limited by 24 hrs/robot/day in the world of atoms. Our new GR00T synthetic data pipeline breaks this barrier in the world of bits. Scaling has been so much fun for LLMs, and it's finally our turn to have fun in robotics! We are creating tools to enable everyone in the ecosystem to scale up with us: - RoboCasa: our generative simulation framework (Yuke Zhu). It's fully open-source! Here you go: http://robocasa.ai - MimicGen: our generative action framework (Ajay Mandlekar). The code is open-source for robot arms, but we will have another version for humanoid and 5-finger hands: https://lnkd.in/gsRArQXy - We are building a state-of-the-art Apple Vision Pro -> humanoid robot "Avatar" stack. Xiaolong Wang group’s open-source libraries laid the foundation: https://lnkd.in/gUYye7yt - Watch Jensen's keynote yesterday. He cannot hide his excitement about Project GR00T and robot foundation models! https://lnkd.in/g3hZteCG Finally, GEAR lab is hiring! We want the best roboticists in the world to join us on this moon-landing mission to solve physical AGI: https://lnkd.in/gTancpNK

Last week, I described four design patterns for AI agentic workflows that I believe will drive significant progress: Reflection, Tool use, Planning and Multi-agent collaboration. Instead of having an LLM generate its final output directly, an agentic workflow prompts the LLM multiple times, giving it opportunities to build step by step to higher-quality output. Here, I'd like to discuss Reflection. It's relatively quick to implement, and I've seen it lead to surprising performance gains. You may have had the experience of prompting ChatGPT/Claude/Gemini, receiving unsatisfactory output, delivering critical feedback to help the LLM improve its response, and then getting a better response. What if you automate the step of delivering critical feedback, so the model automatically criticizes its own output and improves its response? This is the crux of Reflection. Take the task of asking an LLM to write code. We can prompt it to generate the desired code directly to carry out some task X. Then, we can prompt it to reflect on its own output, perhaps as follows: Here’s code intended for task X: [previously generated code] Check the code carefully for correctness, style, and efficiency, and give constructive criticism for how to improve it. Sometimes this causes the LLM to spot problems and come up with constructive suggestions. Next, we can prompt the LLM with context including (i) the previously generated code and (ii) the constructive feedback, and ask it to use the feedback to rewrite the code. This can lead to a better response. Repeating the criticism/rewrite process might yield further improvements. This self-reflection process allows the LLM to spot gaps and improve its output on a variety of tasks including producing code, writing text, and answering questions. And we can go beyond self-reflection by giving the LLM tools that help evaluate its output; for example, running its code through a few unit tests to check whether it generates correct results on test cases or searching the web to double-check text output. Then it can reflect on any errors it found and come up with ideas for improvement. Further, we can implement Reflection using a multi-agent framework. I've found it convenient to create two agents, one prompted to generate good outputs and the other prompted to give constructive criticism of the first agent's output. The resulting discussion between the two agents leads to improved responses. Reflection is a relatively basic type of agentic workflow, but I've been delighted by how much it improved my applications’ results. If you’re interested in learning more about reflection, I recommend: - Self-Refine: Iterative Refinement with Self-Feedback, by Madaan et al. (2023) - Reflexion: Language Agents with Verbal Reinforcement Learning, by Shinn et al. (2023) - CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing, by Gou et al. (2024) [Original text: https://lnkd.in/g4bTuWtU ]

The old approach of sending resumes and hoping for the best isn't working anymore. Thousands of talented engineers are competing for fewer positions. In this market, being skilled isn't enough. You need to be visible. The engineers who are landing roles fast aren't necessarily the most qualified. They're the ones who know how to promote themselves and stand out from the crowd. That's why I created this 5-𝘀𝘁𝗲𝗽 𝗮𝘁𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻 𝘀𝘆𝘀𝘁𝗲𝗺 𝘁𝗼 𝗵𝗲𝗹𝗽 𝘆𝗼𝘂 𝗿𝗶𝘀𝗲 𝗮𝗯𝗼𝘃𝗲 𝘁𝗵𝗲 𝗻𝗼𝗶𝘀𝗲: 📍 Step 1: Optimize Your LinkedIn Profile ↳ Your headline should immediately showcase your specific expertise. ↳ Quantify your achievements. ↳ Make yourself discoverable when recruiters search. 📍 Step 2: Build a Killer GitHub Portfolio ↳ Create 3-4 production-grade projects with detailed READMEs. ↳ Show your thinking process. ↳ Prove your skills instead of just listing them. 📍 Step 3: Write Technical Content Document what you learn. ↳ Share project walkthroughs. ↳ Write about common mistakes. 📍 Step 4: Share Strategically Post your insights with context. ↳ Explain why topics matter. ↳ Document your learning journey consistently. 📍 Step 5: Grow Your Network ↳ Connect with recruiters proactively. ↳ Engage meaningfully with posts daily. ↳ Build relationships before you need them. The result: Instead of competing with hundreds of identical resumes, you become the engineer they already know and want to hire. This system works because it positions you as a known solution, not an unknown candidate. 📌 Want the complete breakdown with actionable tips? Download the full guide here: https://bit.ly/4mZk17A I really hope this is useful. Share this with someone in your network who could benefit from these strategies. 💬 What's the biggest challenge you're facing in this competitive market?

How much do laypersons around the world know about IP? If they know about it, do they have a positive or negative perception of it? And are these changing over time? To answer these important questions which cut right to the heart of popular views and support for IP, we launched WIPO Pulse two years ago – the first ever global survey on IP, covering 50 countries. Now we’ve launched the second edition – this time covering 35,500 laypersons from 74 countries in all regions of the world. The results are interesting and insightful. First, the world is getting savvier about IP. Awareness has grown across all main IP rights since 2023. Copyright and trademarks still lead the pack (no big surprise – music, art, entertainment are fundamental to our lives), but with patents and designs continuing to trail a bit when it comes to public understanding. Second, confidence in the positive impact of IP on the economy remains strong, with two-thirds of respondents (64%) agreeing that IP benefits the economy. Here is where there is a twist – just like in 2023, Asia, Africa and Latin America remain the regions with the most positive perception about IP’s economic benefits, with lower positive perceptions in Western Europe and North America. I welcome your views on this. Third, we were interested in understanding perception among women and youth. Here, we see some gains in awareness among both groups. In Asia-Pacific, awareness rose across all five IP rights for both groups. Western Europe also saw broad gains well. However, youth awareness dipped slightly in Latin America and Eastern Europe. The data we collected is really a wealth of insights that is begging for further investigation. They are valuable not just for WIPO, but the global IP community and local IP institutions, and we will use it to sharpen global, regional and local awareness building, outreach and engagement efforts, as well as combine it with other datasets like the Global Innovation Index to build a deeper picture of the global IP landscape. More: https://lnkd.in/eZ96P-ZJ Photos: WIPO/Berrod #WIPO #IntellectualProperty #Trademark #Patent #Design #Copyright #GeographicalIndications

If you are working in a big tech company on ML projects, chances are you are working on some version of Continuous Integration / Continuous Deployment (CI/CD). It represents a high level of maturity in MLOps with Continuous Training (CT) at the top. This level of automation really helps ML engineers to solely focus on experimenting with new ideas while delegating repetitive tasks to engineering pipelines and minimizing human errors. On a side note, when I was working at Meta, the level of automation was of the highest degree. That was simultaneously fascinating and quite frustrating! I had spent so many years learning how to deal with ML deployment and management that I had learned to like it. I was becoming good at it, and suddenly all that work seemed meaningless as it was abstracted away in some automation. I think this is what many people are feeling when it comes to AutoML: a simple call to a "fit" function seems to replace what took years of work and experience for some people to learn. There are many ways to implement CI/CD/CT for Machine Learning but here is a typical process: - The experimental phase - The ML Engineer wants to test a new idea (let's say a new feature transformation). He modifies the code base to implement the new transformation, trains a model, and validates that the new transformation indeed yields higher performance. The resulting outcome at this point is just a piece of code that needs to be included in the master repo. - Continuous integration - The engineer then creates a Pull Request (PR) that automatically triggers unit testing (like a typical CI process) but also triggers the instantiation of the automated training pipeline to retrain the model, potentially test it through integration tests or test cases and push it to a model registry. There is a manual process for another engineer to validate the PR and performance reading of the new model.  - Continuous deployment - Activating a deployment triggers a canary deployment to make sure the model fits in a serving pipeline and runs an A/B test experiment to test it against the production model. After satisfactory results, we can propose the new model as a replacement for the production one. - Continuous training - as soon as the model enters the model registry, it deteriorates and you might want to activate recurring training right away. For example, each day the model can be further fine-tuned with the new training data of the day, deployed, and the serving pipeline is rerouted to the updated model. The Google Cloud documentation is a good read on the subject: https://lnkd.in/gA4bR77x  https://lnkd.in/g6BjrBvS ---- Receive 50 ML lessons (100 pages) when subscribing to our newsletter: TheAiEdge.io #machinelearning #datascience #artificialintelligence

During my time serving in government, I saw firsthand how geopolitics can impact energy production and flows, with cascading impacts on market and macroeconomic trends.   We're already seeing this play out following the last few days in the Middle East. U.S. and Israeli strikes on Iran triggered retaliatory action across the region that has disrupted energy production and transit.   The market reaction is changing quickly. Since I recorded this video on Monday, oil and gas prices have jumped further, and equities have shifted toward a risk-off move as investors price in continued escalation. Bonds sold off further, reflecting inflation fears in developed markets. Due to the segmented nature of natural gas markets, the impact of higher prices will hit regions differently, with Europe more exposed than the U.S. to elevated LNG prices.   The central question: will this remain a short-term volatility spike or evolve into a broader supply shock? The duration of the disruption and the severity of transit impacts are the core variables I'm watching.   ⬇️ Watch the full video for my latest take on what this could mean for markets.

Quantum computing has officially entered the supply chain. In the newest edition of Supply Chained, I explore why quantum computing is no longer theoretical, abstract, or “someday” technology. After speaking with Murray Thom from D-Wave, one thing became clear: We’ve crossed the threshold from curiosity to capability. This isn’t about physics. It’s about outcomes. ✔ Faster scheduling decisions ✔ Better production plans ✔ Lower energy consumption ✔ Real improvements in manufacturing operations Companies like Pfizer and BASF are already applying quantum optimization to complex problems like job shop scheduling, cutting cycle times, reducing late products, eliminating overtime, and improving throughput without changing physical infrastructure. For supply chain leaders, the key insight is this: Many of the limits we’ve accepted in planning and optimization were not fixed limits. They were computational limits. Quantum computing introduces a new category of processor, alongside CPUs and GPUs, designed specifically for solving hard optimization problems at scale. It’s not a replacement for existing systems. It’s an accelerator for the exact types of challenges that constrain supply chain performance today. This edition breaks down: • What quantum computing really is (in business terms) • Why energy efficiency may matter as much as speed • Where it fits in digital transformation strategies • Why leaders should begin experimenting now If you're serious about the future of supply chain performance, this is a capability worth understanding early. Read the full article in this week’s edition of Supply Chained. ~Mr. Supply Chain® #SupplyChain #SupplyChained #QuantumComputing #DigitalTransformation #AlwaysBeLearning

The frog that halted a dam In Brazil, conservation victories are often framed as heroic struggles against deforestation or mining. In 2014, one involved a toad, reports Thamys Trindade. Melanophryniscus admirabilis, a thumb-sized amphibian found only along a short stretch of the Forqueta River in Rio Grande do Sul, became the decisive factor in stopping a small hydroelectric dam planned less than 300 meters from its habitat. Classified as critically endangered after careful fieldwork, the species forced regulators and prosecutors to accept an inconvenient conclusion: even modest infrastructure can be incompatible with biological survival. That episode is now more than a legal curiosity. In 2024, record floods swept through southern Brazil, submerging the rocky outcrop where the toad breeds and raising doubts about whether the population still existed. When researchers returned in 2025, they found fewer animals than in peak years, but evidence of continued reproduction. Tadpoles were present. Adults had shifted micro-habitats. The system, though altered, had not collapsed. The story carries a lesson with broader relevance. Environmental impact assessments tend to treat extreme climate events as statistical outliers. Yet the National Water and Basic Sanitation Agency projects that floods in southern Brazil could become up to five times more frequent. For species with narrow ecological requirements, resilience depends not on average conditions, but on whether rare refuges persist through shocks. The admirable little toad survives because its habitat was left intact before disaster struck. Had the dam gone ahead, there would have been no margin for recovery. Conservation, in this sense, functioned less as preservation than as risk management. Small species rarely halt big projects in much of the world. When they do, they sometimes reveal why precaution is cheaper than repair. 🐸 English: https://lnkd.in/gRka7EXG 🐸 Portuguese: https://lnkd.in/g22MBPG9

I write. He edits. She approves. They present. And the client still says: “I didn’t get the point.” Welcome to consulting ,where your deck won’t save you. But your clarity will. In a 2024 Bain & Company report, 83% of consultants across strategy, risk, and healthcare roles said communication clarity was more important than technical know-how during client reviews. And a LinkedIn Global Workplace Study found that among consultants aged 22–35, “executive communication” is the #1 skill gap during performance appraisals. Whether you’re a student aiming for BCG, a business analyst at EY, or a healthcare consultant decoding diagnostics for a Tier-2 city hospital, your ability to structure, simplify, and sell your message is what sets you apart. Cheers to our 3 months Leadership Communication program delivered at Deallus for all the senior consultants. Here are my secret beans from our training program : - Minto Pyramid Principle (Think: Top-down thinking) How to use it: ➡ Start with the main recommendation or conclusion. ➡ Back it up with 2–3 grouped arguments. ➡ Use logic and hierarchy to order them. Instead of: “First we did X, then we found Y, hence we suggest Z” Say: “We recommend Z because X and Y indicate…” Bridging Technique (Especially during tough conversations) How to use it: ➡ Acknowledge the question ➡ bridge it to your message ➡ deliver your point. “That’s a valid concern. What we’ve seen across 4 client projects is…” Use this during steering committees, Q&A rounds, or when you’re cornered. Contrast for Clarity (Great for decision-making slides) How to use it: State what something is, followed by what it is not. “This is not just an app upgrade. It’s a workflow redesign that improves patient handover by 40%.” Especially in healthcare consulting — where stakeholders include doctors, government officials, and global NGOs — communication is not a luxury. It’s a lifesaving skill. If you’re leading a consulting team or preparing your analysts for client-facing roles, I design hands-on Leadership Communication Programs to help your team think, write, and speak with executive clarity. DM me or drop a comment — let’s make your team unstoppable. Btw, what’s your way of communicating well in the world of corporate. #training #communication

For the last part of my Energy Resilience series, we have to talk about the worst-case scenario – when the lights actually go out. Earlier this year we saw that happen in Spain and Portugal. A major blackout left millions without power. Trains stopped, shops couldn’t take card payments, hospitals and factories switched to backup. A wake-up call that modern life depends on electricity in ways we often forget until it is gone.   This is what happens when grids are pushed to the edge by fast-moving disturbances or extreme conditions. A couple of years ago, South Australia experienced a state-wide blackout after severe weather took out multiple transmission lines. Investigations showed the system lacked enough inertia to stay stable through the shock. Part of the solution was to install synchronous condensers – giant flywheels that give the grid “weight” and stability. Siemens Energy delivered two of them as part of the response. Not the only measure of course – adapting regulation is also essential – but it showed something important: without resilience in the system, recovery is slow and uncertain. So what do we actually need if we want a fast ramp-up after a major incident? From my perspective, it comes down to three things. 1️⃣ Standardize before the crisis: When parts fail, every minute spent interpreting drawings or debating specifications is a minute the lights stay out. Standard equipment and uniform processes mean teams can move quickly because they are working with tools they already know. Recovery begins long before the fault happens. 2️⃣ Design power plants with failure in mind: A fast restart depends on assets built to recover quickly, not just run efficiently. That means black-start capability, smart redundancy where it matters and systems that can restart without waiting for the wider grid. In the U.S. for example we supported a power plant with a battery system that enables multiple restart attempts within one hour – resilience designed into the plant itself. 3️⃣ No improvisation in the dark: A blackout is the worst moment to negotiate who does what. Good restoration plans spell out which assets come back first, how to stabilize small sections of the grid and when to reconnect them safely. Regular drills with operators, authorities and major customers turn these plans into routine rather than theory. These steps matter because in any major incident skilled people are often the scarcest resource – grid operators, field crews and technical specialists. That is why preparation matters so much. Clear roles, common standards and trusted partnerships mean limited teams can do more in less time. Because when the worst happens what people remember is how long it stayed dark. I hope you have found this mini-series useful. I know social media is often about speed and short takes but sometimes – especially on important topics like this – I find it worthwhile digging into the detail together.✍️ I’d be interested to hear if you agree.