ZML
"Model to Metal"
Paris-based open-source project shipping a hardware-agnostic AI inference compiler stack that runs models on NVIDIA, AMD, TPU, and Trainium chips.
ZML is a Paris-based software company that builds an open-source, production-grade AI inference stack designed to decouple AI model deployment from proprietary hardware. Its core product, also called ZML, compiles machine learning models directly to multiple hardware accelerator targets, including NVIDIA GPUs, AMD GPUs, Google TPUs, AWS Trainium, and AWS Inferentia 2, allowing the same codebase to run across these platforms without requiring separate rewrites for each vendor's software stack. The project is built on the Zig programming language together with the OpenXLA and MLIR compiler infrastructure and the Bazel build system, and it is distributed under the Apache 2.0 open-source license through the public repository at github.com/zml/zml, which has accumulated over 3,500 GitHub stars and more than 700 commits. The company describes its engineering philosophy around three principles: explicit behavior over implicit behavior, composability over monolithic systems, and predictability over 'magic' abstractions, and it markets the product with the tagline 'Model to Metal' and the promise of running 'Any model. Any hardware. Zero compromise.' The toolkit supports popular open model families such as Llama and Qwen, and includes a virtual file system layer that can load model weights from local storage, Hugging Face repositories, HTTPS endpoints, or Amazon S3, alongside built-in sharding and distributed-execution support for scaling inference across multiple accelerators. In 2026 the company released ZML/LLMD, a further inference-serving tool aimed at speeding up AI inference workloads across heterogeneous chip types, which it distributes for free; it also shipped supporting tools such as zml-smi, a monitoring utility, and reported performance improvements including a claimed tenfold speedup in tokenization. ZML maintains a technical blog documenting these releases, operates a public Discord community for developers, and publishes documentation at docs.zml.ai. The company's GitHub organization profile lists its location as France and identifies engineers including a contributor using the handle r-chong among its visible team. ZML positions its target users as engineering teams that need to run large language model and other AI inference workloads in production without being locked into a single hardware vendor's software ecosystem, competing in the space of hardware-agnostic ML compiler and inference-serving tools alongside projects such as vLLM and TensorRT-LLM.
Primary type & automation activities this supplier delivers:
ZML
- Launch Year: 2024
Production inference stack compiling AI models directly to NVIDIA, AMD, Google TPU and AWS Trainium from one codebase.
ZML/LLMD
- Launch Year: 2026
Free self-contained LLM inference server running Llama, Gemma, Qwen and Mistral across 5 chip architectures with DFlash decoding.
Company Facts
Founded
-
Primary Role
Software/Algorithm
Company Size
employees 50-100
Primary Region
Europe
Annual Sales
-
Funding Stage
-
Funding Total
-
iiQKA.AI Vision Enhances Industrial Image Processing with User-Friendly Interface
iiQKA.AI Vision has introduced a standalone vision application within iiQWorks, designed to simplify industrial image processing. This application features a graphical user interface that allows users to implement vision applications efficiently, utilizing drag-and-drop functionality, pre-built templates, and guided workflows from setup to commissioning. The significance of this development lies in its ability to ensure process reliability through robust image processing capabilities, including precise 2D and 3D object recognition, automatic position determination, and gripper point calculation. The system is designed to perform reliably under varying component conditions and changing lighting, making it suitable for diverse industrial applications. Looking ahead, the seamless integration of iiQKA.AI Vision with iiQWorks and the robot controller facilitates a unified system from engineering to operation. This integration supports a consistent workflow from configuration to deployment, along with centralized management and monitoring via a web interface. No further timeline was disclosed at the time of publication.
First Coordinated Drilling Robot Achieves 10x Speed and 99% Accuracy
A groundbreaking drilling robot has been developed, showcasing the ability to perform coordinated operations with a speed ten times greater than traditional methods and an accuracy rate exceeding 99%. This innovation marks a significant advancement in drilling technology, potentially transforming various industrial applications. The introduction of this robot is crucial as it addresses the limitations of conventional drilling techniques, which often suffer from slower speeds and lower precision. By enhancing operational efficiency, this technology could lead to cost savings and improved outcomes in sectors that rely on drilling processes. Looking ahead, industry stakeholders should monitor the deployment and performance of this drilling robot in real-world scenarios. No further timeline was disclosed at the time of publication.
Robotics Automation AI
IROS 2026 Best Paper Finalists: South Korea Leads
Ten papers made the IROS 2026 Best Paper and Best Student Paper shortlist, four from South Korea. What each one does and what it means for industry.
Harmonic Drive to Present on Integrated Actuators at RoboBusiness 2026
At RoboBusiness 2026, Marcel Almonte from Harmonic Drive will lead a session on integrated drive actuator design. Scheduled for October 20 at 10:30 a.m. PT in Santa Clara, California, the session will explore the benefits of integrated actuators in reducing engineering complexity and enhancing system performance. This presentation is significant as it addresses the critical decision-making process in selecting motion architectures for robotics. Attendees will learn how integrated actuators can streamline system integration, reduce cabling complexity, and accelerate time-to-market, ultimately leading to more efficient robotic designs. Participants should watch for practical insights on selecting motion solutions tailored to specific application requirements and constraints. No further timeline was disclosed at the time of publication.
Actuators / Motors / Servos Events Motion Control News Robot Components Technologies
Elmo Motion Control Introduces Innovative Titanium and Platinum Servo Solutions
Elmo Motion Control has unveiled its latest innovations, the Titanium multi-axis motion controller family and the Platinum servo drive series. These products integrate multi-axis control, high computing power, and functional safety into a compact design, enabling machine builders to enhance performance while minimizing component complexity. This development is significant as it addresses the growing demands in sectors like semiconductors, robotics, medical devices, and defense, where traditional machine designs often rely on multiple controllers and external safety systems. Elmo's integrated approach allows for smaller, faster, and safer machines, ultimately reducing the total cost of ownership for OEMs. Looking ahead, the focus will be on how these advancements can further streamline installation and commissioning processes, particularly with the introduction of Elmo’s Application Studio III (EASIII). No further timeline was disclosed at the time of publication.
Transforming Industrial Maintenance: The Role of Agentic and Physical AI in Repair Efficiency
The industrial sector has long relied on predictive maintenance to foresee equipment failures, but this approach often leads to administrative bottlenecks. Alerts generated by AI models trigger slow, manual processes that hinder timely repairs. To overcome these challenges, industries are shifting towards a model that incorporates agentic AI and physical AI, enabling autonomous orchestration of logistics and execution of repairs. Agentic AI allows for the autonomous management of maintenance tasks, optimizing scheduling and logistics without human intervention. This transition is crucial for enhancing operational efficiency, as it transforms maintenance from a reactive process into a self-optimizing engine. For instance, in high-volume manufacturing, agentic AI can immediately address issues flagged by predictive models, significantly reducing downtime and improving return on investment. Looking ahead, the integration of agentic and physical AI is set to redefine maintenance operations. As organizations adopt these technologies, they must ensure safety through deterministic bounding to mitigate risks associated with autonomous decision-making. The future of industrial maintenance will likely see a seamless blend of AI-driven logistics and physical execution, paving the way for more efficient and effective repair processes.
Factory / Analytics
Exploring Custom Motor Solutions and Advanced Motion Control Technologies
Mosrac Motor Co., Ltd. specializes in custom frameless motor solutions tailored to various applications, focusing on torque, voltage, size, structure, and materials. Their engineers collaborate with clients to design motors that meet specific requirements, making Mosrac a reliable partner for high-performance motors. The H-811.I2HS hexapod from PI USA (Physik Instrumente) offers six-axis motion and high dynamics, ideal for precision applications in photonics, optics, and automation. This compact system features patented overload protection, ensuring reliability during operation and setup, which is crucial for demanding environments. Additionally, KNF vacuum pumps and Harmonic Drive's miniature speed reducers are highlighted for their durability and precision in automation applications. With innovations in motor technology and control solutions, companies like MMP Inc. and Elmo Motion Control Inc. are pushing the boundaries of performance and efficiency in robotics and automation systems. No further timeline was disclosed at the time of publication.
Automation Solutions for the Future of Sheet Metal Processing
The increasing variety of parts, shorter cycle times, and a shortage of skilled labor are intensifying the demand for automation in sheet metal processing. Schmalz addresses this need with a comprehensive portfolio of end-of-arm tooling (EOAT) and handling solutions, including vacuum grippers and gripping systems, as well as intelligent complete solutions for automated processes. This shift towards automation is crucial as manufacturers seek to enhance efficiency and productivity in response to market demands. The ability to quickly adapt to diverse production requirements while maintaining quality is essential for competitiveness in the sheet metal industry. Looking ahead, the focus will be on how companies like Schmalz continue to innovate and provide solutions that meet the evolving challenges in sheet metal processing. No further timeline was disclosed at the time of publication.
Allgemein
Teradyne Launches Titan HP Platform with Enhanced Burn-In for AI Data Center Devices
Teradyne, Inc. has unveiled its Titan™ HP platform, which features advanced burn-in capabilities tailored for high-power AI and automotive devices. This innovative system allows for per-device thermal control and realistic workload testing, optimizing reliability stress functions and reducing costs associated with testing and time to market. The introduction of the Titan HP platform is significant as it addresses the increasing demand for high-reliability testing in AI silicon, akin to standards in the automotive sector. By enabling system-level reliability stress testing, Teradyne aims to enhance the quality of shipped devices while managing capital costs and manufacturing space constraints. Looking ahead, Teradyne's Titan HP is positioned to support the evolving power requirements of future AI accelerators. The platform is already available for production use, with demonstrations scheduled at upcoming industry events, including SEMICON West and the IEEE International Test Conference.
Understanding Connector Failures in Robot Cabling: M8 and M12 Selection for Motion Applications
The article discusses the critical factors affecting the performance of M8 and M12 connectors in robotic applications. It emphasizes that while connectors may meet datasheet specifications, their performance in real-world motion scenarios is often overlooked. This oversight can lead to recurring failures in robotic systems, particularly when connectors are subjected to mechanical loads during operation. Understanding the dynamics of robot movement is essential for ensuring reliable connections. The article highlights that connectors and cables must be evaluated as part of a complete moving system, considering factors such as cable length, support points, and the mechanical loads introduced by the robot's motion. Failure to address these issues can result in intermittent faults that persist even after component replacement. Looking ahead, it is crucial for engineers to assess the entire motion envelope of robotic systems when selecting connectors. This includes evaluating the impact of cable routing and support on connector performance. No further timeline was disclosed at the time of publication.
Software, Compute and Interaction at WRC 2026
47 WRC 2026 exhibitors supply robot software, compute or interaction — but only 19 call themselves software companies. Five open artefacts, licence-checked.
WRC 2026 Core Components: 137 Robot Parts Suppliers
The third report in the WRC 2026 series. Report 1 mapped all 248 exhibitors by category and Report 2 covered the 139 that build finished robots. This one covers the 137 that build what goes inside them.
WAIC Review 1/4: Embodied AI's Industrialization Wave: Financing Data, Unit Economics, and a Widening Split
China's embodied AI sector raised $12.5B in H1 2026, but WAIC 2026 speakers detailed ROI demands, power bottlenecks, and the mass-production-to-sales gap.
WAIC Review 4/4: From Port AGVs to Hospital Night Shifts: What WAIC 2026 Reveals About Real Robot Deployments
From 530+ AGVs at a Singapore port to robots on Apple's line, WAIC 2026 speakers shared concrete data on where embodied AI robots are actually working today.
US Restrictions on Chinese Humanoid Robots: The Full Map
Every US restriction on Chinese humanoid robots — NDAA Sec. 163, GUARD Act, the 1260H list — plus the supply-chain math and a 4-tier map of decoupling.
WAIC 2026: Embodied Intelligence Enters Industrial Deployment Era
WAIC 2026 in Shanghai (July 17–20) saw 200+ embodied intelligence companies, 300+ live robots, and first confirmed humanoid deployments in logistics and pharmacy. RobotToday engineering analysis.
Hot French startup ZML releases free product to speed inference across lots of AI chips
ZML, a hot French AI startup endorsed by Turing Award winner Yann LeCun, has now released ZML/LLMD, software that could make running AI less costly.
AI Fundraising Startups AI inference Exclusive ZML
When a Materials Giant Meets Robotics: Covestro at Automate 2026
At Automate 2026, Covestro brought Makrolon, Bayblend, Baysafe BEF and Vulkollan to robotics — tackling weight, heat, sensing and sustainability for humanoids and AGV/AMRs.
The EV Déjà Vu: Why China's Humanoid Robot Blitz May Be Winning the Battle — but Losing the War
China controls 80% of humanoid robot shipments and crushes hardware costs. But is winning the body enough? A deep analysis of the US-China robot war beyond the EV analogy.
ROBOTTODAY Weekly June 23 – 27, 2026
This week in robotics: Agility Robotics $2.5B SPAC deal for first humanoid IPO, BMW Figure 03 logistics deployment, Morgan Stanley doubles China humanoid forecast, ON Semiconductor acquires Synaptics for $7B, Sony halts aibo in Japan. June 23–27, 2026.
