Launching today
Axonnic
Muscle-sensing sleeves for robot training
27 followers
Muscle-sensing sleeves for robot training
27 followers
Axonnic’s forearm sleeve records muscle signals to track finger movement and estimate fingertip force. We combine EMG, video and hand pose into training datasets for robot tasks such as sewing, tool handling and delicate assembly.
Payment Required
Launch Team / Built With
Framer AI AgentsDesign and publish professional sites with AI
Promoted
Hey Product Hunt — I’m Max, co-founder of Axonnic with Arsh. We’re in YC F26, building training data for dexterous robots from human muscle signals.
Our wearable forearm sleeve reads muscle activity to track finger movements and estimate fingertip force, while leaving the hands completely free. We synchronize EMG with video and hand pose so robotics teams can learn from both the visible movement and the muscle activity behind it.
When the hand goes inside the box, camera-based tracking loses sight of it. Axonnic keeps decoding finger movement from forearm EMG. We’ve also demonstrated live control of a robotic hand from those signals.
The tasks we’re targeting include guiding fabric through a sewing machine, handling an electrician’s tools, and positioning small parts during delicate assembly. These are examples of the skilled work we want to turn into robot-training datasets.
Why us? I studied Biology (MBiol) at Oxford, where my research on high-density surface EMG alignment earned the highest research mark in my cohort. I’m now publishing that work as a first-author paper. I saw Skyscanner grow from its early days into a unicorn through to its sale, and also built internal tools there. As an early employee at Voxblock, I worked across PCB development and supply chain.
Arsh studied medicine at Oxford, has published clinical research, and worked on machine learning for cancer and stroke imaging. He also wrote control software for OxRAM’s undergraduate quadruped robot and helped build Medical Advice GPT, which passed 10,000 user interactions. Together, we bring biological signal research, machine learning, robotics and hardware experience.
Sleeves are shipping and paid pilots are underway with our first design partners. We’re opening a few more spots for robotics teams and manufacturers with dexterous tasks they’ve struggled to automate.
What manipulation task would you want to see us tackle next? If you train robot policies, which missing signals most limit your current demonstrations?
As an architecture student who spends way too many hours on delicate model-making by hand, the idea of capturing that kind of fingertip force and turning it into robot training data is fascinating. Sewing and small-part assembly feel like exactly the tasks robots still can't do well. Congrats on shipping real hardware with paid pilots on day one!
Fellow maker launching today, good luck 🚀