AVIN is a subsea edge AI computer that brings artificial intelligence and machine learning to the seabed. Built on an embedded AI computing platform, it acquires, processes, analyses, and stores video, acoustic, and sensor data locally — running computer vision and multi-sensor fusion in real time instead of sending raw data to the surface. An integrated managed PoE switch connects up to eight Gigabit Ethernet devices (cameras, sonars, hydrophones, and other sensors) with IEEE 1588 PTP time synchronisation, while an internal NVMe SSD of up to 10 TB records everything at the edge. The AVIN500 (300 Wh) and AVIN1800 (1 kWh) add an internal rechargeable battery and UPS behaviour for autonomous, unattended deployments on ROVs and AUVs. Every unit runs Linux / Ubuntu with an open software stack (Docker, ROS, TensorFlow, PyTorch, ONNX, OpenCV) in a pressure-rated Aluminium or Titanium enclosure depth-rated to 6000 m.
- Edge AI / ML computingRuns AI and machine-learning models locally for real-time decisions, with no dependence on cloud or surface processing.
- Computer vision & acoustic intelligenceProcesses multiple video and acoustic streams for object detection, classification, tracking, target recognition, and sensor fusion.
- Integrated PoE sensor hubA managed PoE switch connects up to 8 Gigabit devices — cameras, sonars, hydrophones — with IEEE 1588 PTP time synchronisation for deterministic acquisition.
- High-speed edge recordingInternal NVMe SSD up to 10 TB continuously records video, acoustic, and sensor data directly on the seabed.
- Autonomous, battery-backed operationAVIN500 (300 Wh) and AVIN1800 (1 kWh) add an internal rechargeable battery and UPS for unattended AUV / ROV deployments, auto-charged over PoE.
- Edge data reductionProcesses data locally and transmits only alarms, events, metadata, or compressed information to save communication bandwidth.
- Open software architectureLinux / Ubuntu with Docker, ROS, Python, C++, TensorFlow, PyTorch, ONNX Runtime, and OpenCV — deploy your own AI models, optionally GPU-accelerated.
- Subsea-rated platformCombines compute, PoE networking, storage, and battery backup in a pressure-rated Aluminium or Titanium enclosure, depth-rated to 6000 m.
Model range
| AVIN100 | AVIN500 | AVIN1800 | |
|---|---|---|---|
| Internal battery | — | 300 Wh | 1 kWh |
| System memory | 8–64 GB | 8–64 GB | 8–64 GB |
| SSD (NVMe) | 0 / 2 / 10 TB | 0 / 2 / 10 TB | 0 / 2 / 10 TB |
| Ethernet / PoE | 2 / 4 / 8 GbE | 2 / 4 / 8 GbE | 2 / 4 / 8 GbE |
| Time sync | IEEE 1588 PTP | IEEE 1588 PTP | IEEE 1588 PTP |
| Operating system | Linux Ubuntu | Linux Ubuntu | Linux Ubuntu |
| Housing | Aluminium / Titanium | Aluminium / Titanium | Aluminium / Titanium |
| Depth options | 300–6000 m | 300–6000 m | 300–6000 m |
- Type
- Subsea edge AI / ML computer
- Models
- AVIN100 / AVIN500 / AVIN1800
- System memory
- 8–64 GB
- Storage
- NVMe SSD, up to 10 TB
- Internal battery
- 300 Wh (AVIN500) / 1 kWh (AVIN1800)
- Networking
- 2 / 4 / 8 × Gigabit PoE (IEEE 802.3af/at)
- Time sync
- IEEE 1588 PTP
- Operating system
- Linux (Ubuntu)
- Housing
- Aluminium / Titanium
- Depth rating
- 300 / 1000 / 2000 / 6000 m
Specifications subject to change without notice.
FAQ
Subsea edge AI FAQs
What is a subsea edge AI computer?
A subsea edge AI computer runs AI and machine-learning models underwater, processing sensor data locally instead of sending raw data to the surface. AVIN acquires, analyses, and stores video, acoustic, and sensor data at the edge for real-time decisions.
Which sensors can AVIN connect and synchronise?
Up to eight Gigabit Ethernet / PoE devices — cameras, sonars, hydrophones, and other Ethernet sensors — through an integrated managed PoE switch, with IEEE 1588 PTP time synchronisation for deterministic, multi-sensor acquisition.
How much data can AVIN record?
AVIN includes an internal NVMe SSD of up to 10 TB for continuous recording of video, acoustic, and sensor data directly on the seabed.
Can AVIN run autonomously on an AUV or ROV?
Yes. AVIN500 (300 Wh) and AVIN1800 (1 kWh) add an internal rechargeable battery and UPS behaviour for unattended, autonomous operation, with automatic startup and watchdog recovery.
What AI software does AVIN support?
An open Linux / Ubuntu stack with Docker, ROS, Python, C++, TensorFlow, PyTorch, ONNX Runtime, and OpenCV, so you can deploy and update your own AI / ML models — optionally GPU-accelerated.
Also in the lineup
Related products
ANIK Subsea Battery Packs
ANIK subsea battery packs — 100 Wh, 500 Wh and 1.8 kWh lithium-ion power with integrated charging, power-path UPS, and diagnostics, rated to 6000 m.
View product →
LacusOrb Monitoring Buoy
LacusOrb is a self-powered, solar-charged monitoring buoy (150 mm and 300 mm) that streams real-time environmental data to the cloud over satellite or cellular, with a modular multi-sensor payload.
View product →
SAMIN Subsea Smart eFuse
SAMIN subsea smart eFuse — a resettable, programmable, remote-controlled electronic fuse with predictive monitoring for subsea machinery. Up to 850 VDC and 300 A, depth-rated to 4000 m.
View product →
Marinus-DC Subsea Step-Down Converters
Marinus-DC subsea step-down converters take a 400 or 800 VDC bus down to isolated 24 or 48 VDC for auxiliary electronics, control systems, and battery charging — up to 75 A, 4000 V isolation, depth-rated to 6000 m.
View product →
VoltMaris Subsea DC-AC Inverters
VoltMaris subsea DC-AC inverters deliver clean pure-sine AC power from DC battery systems — 115/230 VAC, single-phase and three-phase, 50/60 Hz, depth-rated to 2000 m.
View product →
DeepTorque Subsea Motor Drivers
DeepTorque subsea motor drivers for underwater propulsion, thrusters, pumps, winches and actuators — low-voltage, 400 VDC and 800 VDC, up to 20 kW and 150 Arms, depth-rated to 2000 m.
View product →