Sample price, not a volume price.
Request a sampleA wearable pendant, 7.9 × 4.3 × 1.9 cm and 46 g, with a 4G module and battery inside. It goes online the moment it powers on — no school Wi-Fi, no device-by-device provisioning. A child holds the button, speaks, releases: speech recognition returns in 1–2 seconds and playback starts in 1.1 seconds. Behind it sits a full back office: teachers upload their own teaching material to build a knowledge base, configure the agent their class talks to, then push it to every device in that class. Usage, questions asked and topics not yet mastered all surface in the admin console — and can be built into whatever reports the school needs.
Teachers upload textbooks, lecture notes, school rules and FAQs; the system chunks and indexes them automatically.
PINGOOD TECHNOLOGY CO. LIMITED (trading as AI HW Stack) is a Shenzhen-based AI hardware ODM/OEM. We deliver the complete stack — industrial design and tooling, mainboard, Android firmware, the ASR + LLM + TTS pipeline, the companion app and the cloud backend — as one order.
Technology we build on: Android TV, Azure AI Speech, OpenAI speech-to-text, Alibaba Model Studio, Cloudflare Workers.
Published 2026-09-01 · · Contact: kristy@aihwstack.com
Pairing is the most labour-intensive and failure-prone step in a whole-school rollout. The 4G version is provisioned at the factory and registers itself the moment it powers on. It never touches the campus network, so the IT centre does not have to open ports or whitelist anything, a campus password change does not knock every device offline, and moving between classrooms, campuses or taking one home just works.
Measured on real devices, speech recognition returns a result in 1 to 2 seconds and playback starts speaking at 1.1 seconds. A press interrupts it mid-sentence at any time.
Teachers upload their own textbooks, school rules and history, lecture notes and common questions to the backend, and the system splits and indexes them automatically. Student questions are answered from that library first, traceable back to the source document. Questions it cannot answer are collected automatically so teachers can fill the gap. Content updates take effect immediately with no reflashing.
Yes. A teacher fills in a name, role, speaking style and boundaries in the interface, points the character at the knowledge bases it may read, tries a few lines in the backend to check the tone, then pushes it to a selected class. Every device in that class switches at once.
Learning dashboards by class, by student and by knowledge point, conversations archived per student and searchable, device uptime and data usage, and an automatic report after each lesson. The exact reporting definitions are developed to the school's own terms, and data can be exported or connected over an API.
Masking, retention periods and deletion are supported, and a parental consent template is available. The backend carries the school's name, crest and visual identity, so what people see is the school's own system.
Yes. There are two modes. Full customisation delivers hardware, platform, knowledge base, agents and the data backend together, developed around your teaching scenarios. Hardware-only delivers the 4G device with ASR and TTS already tuned on real units, with the agent and platform interfaces open, so you connect your own model and backend and we never touch your content or data.
7.9 × 4.3 × 1.9 cm and 46 grams, with a 500mAh cell, USB-C charging, a built-in speaker and an optional 3.5mm headphone jack. It can hang from the neck or a schoolbag.
