
Dima Havryliuk
Cofounder & CEO
Built Engram’s EMG acquisition and modeling pipeline. Previously built UserWatch and Kortyx.
Connect with DimaTalk to AI without making a sound.
We’re building a non-invasive wearable for intentional, silent speech. Your words, without the world listening.

Some things are easier said.
Just not out loud.
Imagine capturing a note on the move, asking a question in a shared space, or composing a message with silent speech.
Our goal is a natural, private input for AI, using the subtle movements of intentional silent speech.
“Remember this idea for later.”
We started with facial EMG: measuring the electrical activity of speech muscles. Two prototype generations have shaped what comes next.

Compact EMG headband prototype
85.5%
Four fixed-placement facial EMG channels.
Four fixed-placement EMG channels. The 85.5% result used within-session cross-validation on a predefined phrase set.
This result came from the fixed-placement setup, not the compact headband pictured here. Accuracy fell on the headband. Cross-day, multi-user and open-vocabulary performance are still to be demonstrated.
A useful interface has to work when you put it back on. Our next experiments focus on fresh placements and held-out sessions.
We’re also exploring RF sensing to reduce dependence on skin-contact electrodes. Our focus is recognition and stability across sessions.
Talk research with usBounded phrase classification with fixed-placement facial EMG.
RF feasibility alongside repeatability experiments.
Fresh placements, new sessions, then unseen-sentence input in a useful AI workflow.
What if you could talk to a computer without speaking? We each built silent-speech prototypes independently. Now we’re building Engram Labs together.

Cofounder & CEO
Built Engram’s EMG acquisition and modeling pipeline. Previously built UserWatch and Kortyx.
Connect with Dima
Cofounder & CTO
Built his own silent-speech prototype. Experience spanning custom electronics, firmware and apps.
Connect with BalázsThe 3D model is our industrial design concept for an around-ear interface. It explores the form of a future wearable; fit, electronics and RF sensing are still being developed. The examples on this page illustrate the experience we’re working toward.
No. Silent speech means intentionally articulating words without speaking aloud. Our work focuses on signals associated with those speech movements, not your private thoughts or an internal monologue.
Our early facial EMG experiment reached 85.5% within-session cross-validation on 20 predefined phrases and 400 recordings, using one speaker and four fixed-placement channels. Performance fell when moving to a compact headband. Reliable cross-day, multi-user and open-vocabulary operation remain to be demonstrated.
Our EMG work depends on skin-contact electrodes and consistent placement. We are exploring whether radio-frequency sensing can capture useful speech-related movement with less dependence on electrode contact. RF recognition and repeatability are not yet validated.
Engram is in research and development and is not available to buy. Follow our development updates for prototype progress and future opportunities to participate in supervised testing.
We’re looking for investors and collaborators who want to help turn an early signal into an everyday interface.
Let’s talk dima@engramlabs.co