Humans gain bandwidth. Computers gain touch.
Axonnic reads the electrical signals your muscles already produce and classifies your exact hand position in under 20ms — on-device, no cloud. One band controls games, drones, robots, and any application with an input.
What Axonnic sees
Every time a muscle fires it sends an electrical signal. Axonnic reads dozens simultaneously — building a precise, real-time picture of exactly what your hand is doing.
Illustrative signal. Real data streams from a wearable electrode array on your forearm.
It knows what your hand is doing.
Axonnic identifies your exact hand position in under 20 milliseconds — faster than a keypress, more expressive than any button.
PREDICTED CLASS
Fist
91.0% confidence
CLASS PROBABILITIES
Hover to pause · Tap to pause/resume · Cycles automatically · Probabilities are illustrative
What works today.
Two working demos. A pro in the loop. The hardware already works.
Live demos as posted on LinkedIn — unedited hardware footage.
What it controls
One band. Anything with an input.
The same gesture engine drives all of them. Pick a vertical — the input is solved.
Games & esports
A layer on top of existing interfaces. Sub-millisecond muscle signals beat mouse and key — bind full combos to a single flex. Bandwidth is the moat: every other input device tops out; this one scales with your muscles.
Drones & defence
Hands on the weapon. Eyes up. Still in command. Direct a drone flight with muscle signals, drive ground robots under full kit, issue commands silently — latency and free hands decide the engagement.
Robotics & training data
Every session is a labelled dataset of human dexterity. Robotics is bottlenecked on dexterity data — Axonnic generates it as a byproduct of every user, teaching machines the missing sense: touch.
One interface. Infinite applications.
When you can read the exact position of every finger in real time, almost every interaction with a machine becomes possible.
Replace the keyboard
Small finger movements become keystrokes, commands, shortcuts, and gestures. No physical keys. No latency. No hardware in the way of your ideas.
Use it anywhere
No desk, no mouse, no screen required. Axonnic works in the field, on the move, in the operating room — wherever your arm goes, the interface goes.
Build anything
An open platform for developers. Map any gesture to any action. Create controllers that don't exist yet. Test interaction paradigms at the speed of thought.
Tailored to your body
Axonnic learns your specific muscle patterns. The more it knows you, the more precise it gets — personalised accuracy that off-the-shelf hardware can never match.
Teach machines touch
Machines mastered language and vision — touch is the missing sense. Every Axonnic session is a labelled dataset of human dexterity, and the data to teach it comes off the forearm.
Privacy by default
On-device ONNX inference. No cloud dependency, no data transmission. Your muscle signals never leave your hardware — clinical-grade privacy without compromise.
How it works
Three steps. Zero middlemen.
A wearable array reads the signals your muscles already produce. On-device AI interprets them instantly. Your device responds — no cloud, no lag.
Muscle fires
Electrode array reads HD-sEMG across 32–100+ channels at the forearm surface. Every signal, every fibre.
Sensor streams
A wearable sensor array streams raw muscle signals wirelessly to the Axonnic runtime — no cloud hop, no proprietary lock-in.
AI predicts
On-device ONNX inference classifies movement intent in under 20ms with personalised accuracy trained to your exact muscle patterns.
Outputs
The Opportunity
Bioelectric computing is a platform shift.
The wearable input market is a $10B+ category. The interface layer — translating biological intent to digital action — is still wide open. Axonnic is building the infrastructure: land in gaming, expand to every wrist.
Three curves crossed simultaneously: dry-electrode arrays that cost £10,000 per channel now cost under £50; Meta’s $1B CTRL-Labs acquisition validated the category but placed the bet on AR glasses on a five-year horizon, leaving the consumer and gaming window open; and transformer-based processing turned what required a 2019 PhD thesis into a fine-tune. Unlike existing products limited to wrist-snap gestures, Axonnic reads the whole forearm — 32 to 100+ channels — resolving individual fingers, not just gross movements.
The evolution of control
Each new interface unlocked a new class of software. Axonnic is the next one.
$10B+
Total addressable market — forearm input by 2030
$2.5B
Serviceable market — gaming is the innovation hub
$1B
Meta's CTRL-Labs acquisition — category validated
32-ch HD-sEMG
Axonnic Runtime
Our Story
A mission older than computing.
This is not about input devices. It is about where information systems go next — and why we believe the next inflection belongs to the neuromuscular interface.
For most of life on Earth, information was stored and transmitted biologically. DNA and RNA formed a base-4 system capable of preserving complexity across generations. Nervous systems emerged on top — a faster electrical layer capable of learning within a single lifetime.
Humanity then externalised this process: computers, the internet, and now artificial intelligence. External computation is beginning to evolve faster than human interaction with it can keep pace.
We still communicate with machines through interfaces designed decades ago. The bottleneck is no longer computation. It is the translation layer between biological intent and digital action. Axonnic is an intermediate step toward closing that gap — non-invasive, immediate, and grounded in the body's own electrical language.
“The compute is here. The bottleneck is how fast we can tell it what to do.”
That’s what Axonnic fixes.
We are not waiting for neural implants. The body already speaks electrically. We just need to listen.
The Team
Science meets tech.
Maxim Williams
CEO & Co-Founder
Oxford MBiol research — signal processing · EMG calibration paper in preparation (1st author)
Built with a hardware company from the ground up · Seen a unicorn scale from the early days
Arsh Patankar
CTO & Co-Founder
Oxford medic — research in machine learning and AI in cancer and stroke imaging
Published multiple papers · Repeat founder — medical LLM with 300,000+ uses · Worked on Oxford's first quadruped robot
FAQ
Common questions.
Early Access
Be first.
Join researchers, developers, and early builders shaping the next input paradigm.
No spam. Early access invites and product updates only.
The interface of the future
is already part of you.
Every computer interaction you have ever had was constrained by the hardware in front of you. Axonnic removes that constraint.
Talk to the team
We're building the infrastructure layer for bioelectric computing. If you're investing in consumer hardware, HCI, gaming peripherals, or next-generation input paradigms — we'd like to talk, whether you're writing first checks or leading rounds.