MT200 AI Box
Industrial edge AI box built on Qualcomm QCS9075. Up to 200 TOPS, dual NPU, 12–36 GB LPDDR5 and a full industrial I/O set. Runs 7B models locally on Ubuntu.
View specificationsTailAether delivers industrial edge AI systems for factories, vehicles and connected homes — local inference, offline operation, and data that never leaves your premises.
Three building blocks for on-premise intelligence: high-performance edge inference, BIOS-level remote control, and low-power sensor/relay control.
Industrial edge AI box built on Qualcomm QCS9075. Up to 200 TOPS, dual NPU, 12–36 GB LPDDR5 and a full industrial I/O set. Runs 7B models locally on Ubuntu.
View specificationsIP-KVM appliance for BIOS-level remote access. Power, keyboard, video and virtual media over the network — control any machine as if you were at the desk.
PreviewLow-power Wi-Fi and BLE controller for relays, sensors and legacy equipment. Simple HTTP/MQTT API, local automation rules and OTA updates.
PreviewFor factories, clinics and vehicles, sending raw data to a public cloud is often the blocker. Running the model on-site removes it.
Camera feeds, documents and sensor data are processed locally. Nothing is uploaded, which simplifies privacy and compliance reviews.
Inference continues when the network drops — mines, ships, remote sites and factory floors included.
No round trip to a data centre. Local response times stay predictable under load.
One hardware purchase instead of a metered API bill that grows with every query.
Ubuntu 24.04, ONNX, TensorFlow Lite, QNN and SNPE. Bring your own models and tooling.
CAN-FD, PCIe, GPIO, dual Ethernet and a Raspberry Pi 40-pin header for real-world integration.
Validated scenarios, from AI agents that operate hardware to fully offline document workflows.
Send a schematic screenshot to a chat app; the local vision model identifies the pin and drives GPIO to change an indicator light. No cloud hop.
Parse trip records, fill every field of an expense form and attach receipts automatically — work that used to take half an hour per submission.
Cameras feed a local vision model that detects anomalies, then an LLM decides whether to speak up or alert family members. Video never leaves the home.
Competitor research, copywriting and multi-platform publishing run as local agents, keeping strategy and operational data inside the company network.
MT200 is designed and manufactured by MeiG Smart Technology (SZSE: 002881). TailAether provides product selection, integration support and delivery. The MT200 name and specifications belong to MeiG Smart; we help you evaluate, deploy and support it in your application.
Send your use case, model size and I/O requirements — we will reply with a configuration and a quote.
Tell us about your project: workload, model, environment and timeline. We usually reply within one business day.