Edge MLIn production
AI predictive maintenance
On-device anomaly detection that flags failing bearings, motors, and pumps 9–14 days before they fail. Runs on a 32-bit MCU.
The lab
Research that ships.
A portion of every project funds R&D. These are the open questions our engineers are chasing — and what they've already turned into product.
Spotlight
Three things in production.
Edge ML
ConnectivityShipping in v4.1
LoRaWAN protocol stack
Our own stack — 30% lower power, native mesh fallback, and OTA-updatable in place. Used across every Argynix gateway.
Consumer AIPilot · 2026
Self-learning home
An adaptive automation layer that learns occupant rhythms in two weeks. No app-tuning required.
Research areas
What we're chasing next.
Six open questions our team is actively working on. Some become product. Some become papers. All compound.
R/01 · 06
Edge ML & TinyML
Sub-100mW inference for sensors that ship in the wild for 10+ years.
R/02 · 06
Mesh & ad-hoc networks
Self-healing topologies for industrial and rural deployments.
R/03 · 06
Computer vision
Industrial-grade vision pipelines: defect detection, OCR, motion analysis.
R/04 · 06
Industrial protocols
OPC-UA, Modbus, BACnet — bridged into modern observability.
R/05 · 06
Robotics control
Reinforcement-learning loops for pick-and-place and autonomous vehicles.
R/06 · 06
Power & energy
Solar harvesting, battery-less devices, and ultra-low-power architectures.
Collaborate
Working on something hard?
We partner with universities, labs, and operators on early-stage R&D. Drop us a line — we love a hard brief.
Connected · 42 nodes
Edge ML running
99.9% uptime