Edge AI, end-to-end
Language Models, Computer Vision, and Privacy on One Device
Most edge-AI signage products are partial: hardware vendors sell boxes, CMS vendors sell software on someone else's boxes, and specialty AI players sell single-vertical apps. TelemetryOS ships the complete stack — validated AI silicon, the OS underneath, the runtime catalog above it, and the cloud that manages all of it as one fleet.
- On-Device LLMs
- Run open-weight language models directly on the Node Max NPU, iGPU, and CPU. Conversational concierge kiosks, multilingual wayfinding, and retrieval-augmented product discovery answer in real time — without the latency, privacy exposure, or per-query cost of cloud LLMs, and they keep working when the network doesn't.
- Multi-Stream Computer Vision
- Detection, classification, OCR, and embedding models run on live USB or IP camera feeds — visual inspection at line speed, out-of-stock detection on the shelf, queue measurement, occupancy, and audience-aware content. Footage is analyzed on the device and discarded.
- Private by Design
- No cloud round-trip and no PII stored. Inference runs inside the hardened TelemetryOS Edge environment on the device, which is what makes AI deployable where patron and patient data cannot leave the property — healthcare, gaming, education, and other regulated spaces.
How Edge AI runs on TelemetryOS
Container-Managed AI on AMD Ryzen AI Silicon
Every AI workload ships the same way: as a Docker container running beside the TelemetryOS application. The application is the user-facing screen experience; the AI runtime is a sidecar exposing local inference endpoints. TelemetryOS Studio manages both as one fleet.
- AMD Ryzen AI silicon on Node Max. A 12-core Zen 5 CPU, Radeon 890M GPU, and XDNA 2 NPU rated at up to 50 TOPS — up to 80 TOPS of combined platform AI compute. Quad 8K-capable output, up to 64 GB DDR5, and up to 8 TB NVMe give models, RAG indexes, and vision-pipeline state room to live next to the screen application.
- AI runtimes ship as Docker sidecars. llama.cpp with Vulkan acceleration, Ollama as a hosted endpoint, ONNX Runtime for vision and speech, and Whisper for speech recognition — each packaged as a container running beside the TelemetryOS application, exposing local OpenAI-compatible or gRPC endpoints.
- A validated runtime catalog, not a science project. TelemetryOS validates the runtime configurations end-to-end on TelemetryOS Edge and the Ryzen AI silicon, so partners and customers don't build the inference stack from scratch. Specific models refresh with the ecosystem; the runtimes and the silicon are the stable interface.
- OCuLink eGPU expansion when you need more. Node Max exposes four lanes of PCIe 4.0 over OCuLink to an external GPU enclosure — scale a deployment from the integrated NPU and GPU to discrete-GPU class inference without changing the platform underneath.
Regulated and Managed
- Built for PHI and gaming-floor rules. In most regulated jurisdictions, patient data and casino surveillance footage cannot transit cloud AI services. Node Max runs the language model and the vision pipeline on the device inside TelemetryOS Edge — the data never leaves the property.
- Compliance posture that carries through. TelemetryOS is SOC 2 Type I certified and GDPR compliant, with application sandboxing, audit logging, and cloud data residency across US, EU, UK, Canada, and Asia-Pacific regions.
Fleet operations
AI Workloads, Managed Like Everything Else
Edge AI doesn't get its own console, its own update mechanism, or its own deployment scripts. The inference sidecar, the model weights, and the screen application in front of them all flow through the same TelemetryOS pipeline your operations team already runs.
- One Console for Screens and AI
- TelemetryOS Studio manages the AI sidecar and the screen application as one fleet — live device health, remote diagnostics, group deployments, and staged rollouts cover both.
- Signed OTA for AI Containers
- AI container images update through the same signed over-the-air pipeline as the OS and applications, with staged rollouts and one-click rollback across the fleet.
- Git-to-Screen CI/CD
- The screen application that fronts your AI workload deploys the same way as any TelemetryOS app — push a commit and the platform builds, hosts, and deploys it to every assigned device.
- Standard Stacks, No Lock-In
- x86, Linux, Docker, and open-source ML runtimes. Workloads built for Node Max run on the same code paths in cloud development and other Linux x86 environments — no proprietary inference SDK forks.
What teams deploy
From Concierge Conversations to the Production Line
The same platform powers a hotel kiosk answering guest questions at midnight and a vision station scoring parts at line speed. Match the workload to the touchpoint: Node Mini and Node Pro carry the signage, Node Max carries the intelligence — one fleet, one console.
- AI Concierge Kiosks
- A guest asks where the spa is or what time the restaurant closes, and the kiosk answers in their own language — around the clock, with the conversation processed entirely on the device. No PII leaves the property, and the concierge keeps answering through an internet outage.
- AI Visual Inspection
- Defect-detection models analyze parts at line speed with millisecond latency, flag anomalies to the adjacent screen, and log results locally. The same Node Max can correlate defect spikes with MQTT sensor data and surface the trend on the supervisor's wallboard.
- Smart Drive-Through Menus
- The board recognizes a returning vehicle and recalls past orders entirely on the device — no faces, no PII, no cloud round-trip — then greets it with tailored suggestions that lift average ticket size.
- Smart Cooler & Shelf Screens
- A camera over the shelf flags an out-of-stock facing to the store team while the glass keeps selling, and anonymous vision measures dwell and engagement — with nothing identifiable retained.
- Healthcare Digital Front Door
- A multilingual check-in assistant guides patients through arrival and translates on the spot — on-device, because in regulated jurisdictions patient data can't transit cloud LLMs. PHI stays in the facility by architecture, not by policy.
- Multi-Container Edge Compute
- Run the TelemetryOS application, an inference sidecar, an MQTT broker, a content cache, and protocol adapters concurrently. The 12-core Zen 5 CPU and 64 GB DDR5 ceiling carry several heavy containers without contending with display rendering.
How TelemetryOS differs
Edge AI Without the Tradeoffs
Raw TOPS without a managed platform is a hardware project. A CMS with a bolted-on AI partner is an integration project. TelemetryOS Edge AI is one product: the silicon, the hardened OS, the validated runtimes, and the fleet console that operates them together.
- Vertically Integrated
- Hardware (Node Max), OS (TelemetryOS Edge), runtime, SDK, CMS, fleet management, and the AI runtime catalog ship as one product from one vendor — no stitching boxes, software, and AI partners together.
- No CUDA Lock-In
- Standard x86, Linux, Docker, and open-source ML stacks. The same workloads run in cloud development environments and on the device — no proprietary inference SDK to fork around.
- OCuLink eGPU Expansion
- Four lanes of PCIe 4.0 to an external GPU enclosure scale Node Max from the integrated 50 TOPS NPU to discrete-GPU class when a workload demands it.
- Validated AI Runtime Catalog
- llama.cpp, Ollama, ONNX Runtime, and Whisper — validated end-to-end on TelemetryOS Edge with the Ryzen AI silicon. We're clear about what works in production today and what's on the roadmap.
- Regulated-Vertical Ready
- On-device inference satisfies PHI residency requirements in healthcare and data-residency rules on gaming floors. TelemetryOS Edge is the hardened OS that carries the architecture; the platform is SOC 2 Type I certified and GDPR compliant.
- Container-Native AI Delivery
- Every AI workload — inference engine, model weights, RAG index, supporting services — ships as Docker images managed by TelemetryOS Studio, updated and rolled back through the same pipeline as the application.
Explore Edge AI on TelemetryOS
Node Max →
The edge-AI appliance — 50 TOPS NPU, up to 64 GB DDR5, OCuLink eGPU expansion.
AI Concierge Kiosks →
On-device assistants that answer guests in their own language, around the clock.
AI Visual Inspection →
Defect detection at line speed, with results flagged to the floor screens.
Smart Drive-Through Menus →
Anonymous vehicle recognition and tailored promos — no PII, no cloud round-trip.
Smart Cooler Screens →
Cooler-door ads, live planograms, and on-device out-of-stock detection.
Security & Compliance →
SOC 2 Type I, GDPR, hardened OS, and the architecture regulated buyers require.
Bring Edge AI to Your Screens
Talk to our sales and solutions team about Edge AI workloads, Node Max configurations, and validated runtimes for your deployment — or see pricing to plan the fleet around it.