Growing the Robotics Application Ecosystem

𝗔𝗳𝘁𝗲𝗿 𝟭𝟬+ 𝘆𝗲𝗮𝗿𝘀 𝗶𝗻 𝗿𝗼𝗯𝗼𝘁 𝗹𝗲𝗮𝗿𝗻𝗶𝗻𝗴, from my PhD at Imperial to Berkeley to building the Dyson Robot Learning Lab, one frustration kept hitting me: 𝗪𝗵𝘆 𝗱𝗼 𝗜 𝗵𝗮𝘃𝗲 𝘁𝗼 𝗿𝗲𝗯𝘂𝗶𝗹𝗱 𝘁𝗵𝗲 𝘀𝗮𝗺𝗲 𝗶𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲 𝗼𝘃𝗲𝗿 𝗮𝗻𝗱 𝗼𝘃𝗲𝗿 𝗮𝗴𝗮𝗶𝗻? 𝗧𝗵𝗲 𝗽𝗮𝘁𝘁𝗲𝗿𝗻 𝗜 𝗸𝗲𝗽𝘁 𝘀𝗲𝗲𝗶𝗻𝗴:   • New robotics team starts   • Spends 6 months building data collection pipeline   • Spends another 3 months debugging synchronization issues   • Finally starts collecting task-specific data   • Realizes their infrastructure choices limit their flexibility   • Starts over 𝗧𝗵𝗶𝘀 𝗶𝘀 𝘁𝗵𝗲 𝘄𝗵𝗼𝗹𝗲 𝗽𝗼𝗶𝗻𝘁 𝗼𝗳 𝗿𝗼𝗯𝗼𝘁 𝗹𝗲𝗮𝗿𝗻𝗶𝗻𝗴: Robot learning is fundamentally data-driven. Whether you're picking strawberries or assembling electronics, the core infrastructure needs are identical. That's actually why I was so interested in pursuing data-driven robotics over a decade ago. 𝗬𝗼𝘂 𝗮𝗹𝘄𝗮𝘆𝘀 𝗻𝗲𝗲𝗱:   • Multi-sensor data synchronization across different frequencies   • Flexible storage that works with future algorithms   • Visualization tools to understand your data   • The ability to experiment with different temporal resolutions   • Robust logging that captures everything you might need later The trend towards AI in robotics is growing, with robots needing to process and analyze large amounts of sensor data to manage variability and unpredictability in real environments. 𝗕𝘂𝘁 𝗲𝘃𝗲𝗿𝘆 𝘁𝗲𝗮𝗺 𝗯𝘂𝗶𝗹𝗱𝘀 𝘁𝗵𝗶𝘀 𝗳𝗿𝗼𝗺 𝘀𝗰𝗿𝗮𝘁𝗰𝗵. Imagine if every web developer had to build their own database, web server, and deployment pipeline before writing their first line of application code. 𝗧𝗵𝗶𝘀 𝗶𝘀 𝘄𝗵𝘆 𝗜 𝗳𝗼𝘂𝗻𝗱𝗲𝗱 𝗡𝗲𝘂𝗿𝗮𝗰𝗼𝗿𝗲. Instead of every robotics team spending months on infrastructure, we provide the common tools that let you go from "I have a robot" to "I'm shipping intelligent robot behaviors" in days, not months. 𝗧𝗵𝗲 𝗿𝗲𝗮𝗹 𝗶𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 𝗶𝗻 𝗿𝗼𝗯𝗼𝘁𝗶𝗰𝘀 𝘄𝗼𝗻'𝘁 𝗰𝗼𝗺𝗲 𝗳𝗿𝗼𝗺 𝗲𝘃𝗲𝗿𝘆𝗼𝗻𝗲 𝗿𝗲𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝘁𝗵𝗲 𝘀𝗮𝗺𝗲 𝗽𝗹𝘂𝗺𝗯𝗶𝗻𝗴. 𝗜𝘁'𝗹𝗹 𝗰𝗼𝗺𝗲 𝗳𝗿𝗼𝗺 𝘁𝗲𝗮𝗺𝘀 𝘄𝗵𝗼 𝗰𝗮𝗻 𝗳𝗼𝗰𝘂𝘀 𝗲𝗻𝘁𝗶𝗿𝗲𝗹𝘆 𝗼𝗻 𝘄𝗵𝗮𝘁 𝗺𝗮𝗸𝗲𝘀 𝘁𝗵𝗲𝗶𝗿 𝗮𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝘂𝗻𝗶𝗾𝘂𝗲. Robot learning shouldn't be bottlenecked by infrastructure. It should be bottlenecked by creativity. What's the longest you've spent building infrastructure before getting to the actual robotics problem you wanted to solve?

Shoutout to Talia Goldberg/Bessemer Venture Partners for their excellent deep dive into the state of Physical AI, and including Miru (YC S24) alongside other great companies building the infra layer for robotics. They aptly pointed out what's catalyzing growth: 1) The cost of robot hardware is falling exponentially 2) On-device compute is getting both more powerful and efficient (thank you, NVIDIA Jetson!) 3) More research than ever in robot learning Where I disagree is the framing of a potential “ChatGPT moment” for robotics. I don't think there will be one. The unlock for ChatGPT was RLHFing an existing model (GPT-3.5) into a new application (chatbot). In robotics, the problem isn't that we haven't found the right applications for Physical AI. They are abundant and obvious: in logistics, construction, transportation, etc. We have the right applications, but we don't yet have our version of GPT-3.5 because: 1) We don't have enough quality data 2) Our current end-to-end models aren't reliable and long-horizon enough (though GEN-0 may prove that #1 is the more critical limiting factor). How do we solve this? These problems are solved through a bunch of schlep: real-world robot deployments, collecting more data, accounting for corner cases, in-built redundancy, etc. There will be step-change improvements to be sure, but my guess is that the journey will resemble self-driving's decade-long grind to achieve 9s of reliability. Infrastructure will always matter. Regardless, whether it's a fast takeoff or a slow grind, infrastructure tools like Miru will be invaluable in helping robotics teams scale. Just ask the big LLM labs that still struggle with deployment, inference, and monitoring at scale. In the same vein, Physical AI teams of the future will still face challenges in provisioning robots, deploying software, and monitoring their applications. At Miru, we're building these developer tools to help robotics teams scale to millions and eventually billions of robots in their fleets.

🚀 The Evolution of Supply Chain Automation: From Single-Use to an AI-Powered Ecosystem Robotics and automation in the supply chain are no longer just isolated tools—they now form a dynamic, interconnected ecosystem. To stay ahead, businesses must move beyond single-use automation and adopt AI-driven orchestration to optimize how these systems work together. At Kenco Group, we’re leading this transformation by working with strategic partners like GreyOrange and KPI Solutions to develop advanced orchestration software. This technology ensures that autonomous mobile robots (AMRs), robotic picking systems, and AI-driven logistics platforms operate seamlessly, intelligently, and efficiently across the entire supply chain. 📦 Why AI-powered orchestration is essential: ✅ Optimized Robotics Coordination – Ensures different automation systems work in sync ✅ Real-Time Decision Making – AI adapts to demand fluctuations and disruptions ✅ Increased Efficiency & Resilience – Maximizes throughput while reducing errors The future isn’t just about deploying more robots—it’s about intelligently managing automation to drive supply chain agility. With AI at the core, companies can unlock the full potential of robotics and automation to stay competitive in an ever-changing logistics landscape. Excited about where supply chain automation is headed? Let’s discuss! 👇 #AI #Automation #SupplyChain #Robotics #Logistics #SmartSupplyChain #Industry40 #DigitalTransformation #GreyOrange #KPI #Kenco

ROBOTICS ARCHITECTURES Robotics architecture forms the backbone of modern automation, combining physical components and intelligent software to create systems capable of sensing, processing, and acting in real-world environments. It includes both technical architecture—the hardware, control systems, and software platforms that make a robot function—and solution architecture, which ensures that robotics align with business goals and workflows Technical architecture focuses on how sensors, motors, controllers, and computing units work together to perform tasks like movement, perception, and interaction. It also includes software layers that manage real-time decision-making, data processing, and communication between components On the other hand, solution architecture addresses how robots are integrated into broader systems—such as warehouses, hospitals, farms, or retail stores—and how they help achieve operational efficiency, accuracy, or cost savings. This includes designing systems that are scalable, easy to maintain, and responsive to changing business needs In practice, a robotics solution might include autonomous vehicles for moving goods, robotic arms for assembling products, drones for surveying land, or service robots that interact with people. These applications require careful planning to ensure they function smoothly with other technologies like inventory systems, management dashboards, or cloud platforms. A well-planned robotics architecture allows data to flow seamlessly between physical devices and business systems, enabling better monitoring, control, and analysis. It must also support both immediate actions (processed locally on the robot) and more complex decisions (which might use cloud computing) BUSINESS USAGE As the demand for automation grows, businesses are adopting robotics architectures that support learning, adaptation, and continuous improvement. These systems are designed not just to work efficiently, but to evolve over time—learning from their environment, updating their behaviors, and improving performance Whether it’s improving production speed in factories, enhancing patient care in hospitals, or reducing manual labor on farms, robotics architecture is central to building reliable, intelligent, and flexible automation solutions. Businesses also use robotics to lower operational costs, boost accuracy, ensure 24/7 uptime, and gain competitive advantage through smarter, data-driven workflows Industries such as logistics, retail, construction, defense, education, and even entertainment are increasingly adopting robotics architecture to streamline operations, improve safety, personalize services, and handle complex tasks that require precision and consistency FEATURES - Modularity - Scalability - Real-time - Sensor fusion - Autonomy - Adaptability - Interoperability - Fault tolerance - Cloud connectivity - AI integration IMAGE CREDIT https://lnkd.in/ezYSnypk

It’s becoming a little easier to build sophisticated robotics projects at home. Earlier this week, AI dev platform Hugging Face released an open AI model for robotics called SmolVLA. Trained on “compatibly licensed,” community-shared datasets, SmolVLA outperforms much larger models for robotics in both virtual and real-world environments, Hugging Face claims. “SmolVLA aims to democratize access to vision-language-action [VLA] models and accelerate research toward generalist robotic agents,” writes Hugging Face in a blog post. “SmolVLA is not only a lightweight yet capable model, but also a method for training and evaluating generalist robotics [technologies].” SmolVLA is a part of Hugging Face’s rapidly expanding effort to establish an ecosystem of low-cost robotics hardware and software. Last year, the company launched LeRobot, a collection of robotics-focused models, datasets, and tools. More recently, Hugging Face acquired Pollen Robotics, a robotics startup based in France, and unveiled several inexpensive robotics systems, including humanoids, for purchase. SmolVLA, which is 450 million parameters in size, was trained on data from LeRobot Community Datasets, specially marked robotics datasets shared on Hugging Face’s AI development platform.