Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up
Appearance settings

Sipsa Labs

The efficiency layer for AI. Run any model on any hardware - near-losslessly, with verifiable reconstruction. Flagship: UltraCompress.

Sipsa Labs

Embedded intelligence for machines. We build edge-AI that lives inside the machine itself — nodes that perceive, understand, and act, with no cloud in the loop.

Sentio — chapter one

A small, passive sensing node for the drones radios can't hear. A new class of FPV drone flies its control link over optical fiber — zero RF emissions — so the RF-based detection most defenses rely on never sees it. Propulsion, however, can't be muted.

Sentio pairs the cue that can't be turned off with the sensor that confirms it:

  • Hear — 360° passive acoustic cue, no emissions, day or night
  • Confirm — a camera interrogates the cued bearing: drone or not-drone
  • Act — fused tracks publish in standard formats (TAK-compatible) to systems you already run
  • A detection layer, not a weapon — and the whole loop runs on-device, on commodity edge silicon

Where it actually stands

Working prototype runs end-to-end on real edge hardware today. Detection behavior is validated in physics-based simulation — labeled SIM, because simulation is evidence of promise, not performance. The field campaign is the next phase; its numbers will be published with collection methodology, whether they flatter us or not. Until then you will not find a detection-accuracy claim from this company anywhere.

The arc

Defense is chapter one because the need is urgent and specific — not because it's the whole book. The same hear-look-decide loop, retrained, is a node that hears a failing bearing in a motor, an intrusion at a fence line, a pest in a field. Machines with senses and an on-device brain, one fielded mission at a time.

Prior era — archived in public

From 2025–26 Sipsa Labs built UltraCompress (near-lossless LLM weight compression, 23 verified architectures). It was discontinued in June 2026; the research, methodology, and verified results remain public at sipsalabs.com/research/compression-archive — that's what "built in the open" means when a chapter ends. (BUSL-1.1, patents pending.)


Talk to us: founder@sipsalabs.com · sipsalabs.com · /sentio · careers

Popular repositories Loading

  1. ultracompress ultracompress Public

    Verified AI infrastructure for regulated deployment. UltraCompress (our wedge): near-lossless 5-bit compression with SHA-256-reproducible reconstruction - prove the model in production is the one y...

    Python 13

  2. .github .github Public

    Sipsa Labs organization profile

Repositories

Loading
Type
Select type
Language
Select language
Sort
Select order
Showing 2 of 2 repositories

People

This organization has no public members. You must be a member to see who’s a part of this organization.

Top languages

Loading...

Most used topics

Loading...

AltStyle によって変換されたページ (->オリジナル) /