1. The head restaurant track: digitizing the chef's craft
The head restaurant's chef puts on a Galaxy XR and cooks as usual while it records from the chef's own point of view. The AI learns from that footage and builds a standard model of how that head restaurant cooks, unique to it.
- Invitation, then partnership: Korean head restaurants hear about us through a QR code sent by registered mail or a personal introduction, and then move on to a partnership.
- Structured filming: 30–50 hours of head restaurant footage are filmed in stages, with filming guides by time of day, cooking step and ingredient.
- Automatic quality scoring: the AI scores the quality of every clip, keeping both the amount and the quality of training data in check.
- AI training: the footage builds up into a model unique to that head restaurant (Multi-LoRA fine-tuning).
2. The learner (franchise) track: one-to-one comparison with the head restaurant
The franchise chef wears a Galaxy XR while cooking the same dishes as the head restaurant, and the AI compares each movement with the head restaurant's standard in real time.
- Theory: before hands-on practice, the AI talks the learner through the head restaurant chef's know-how, ingredients and tools by voice.
- Practice: the AI compares the learner's movements with the head restaurant's every 0.1 seconds and scores the gaps in four areas: motion, timing, ingredients and environment.
- Regular refreshers: learning doesn't end after one round. Small gaps from the head restaurant that can creep in over time are closed through regular retraining.
3. What "our AI" actually is
"Our AI" isn't a single model. It's an integrated system.
- Gemini 2.5 Flash: a multimodal LLM fine-tuned on head restaurant footage (the core for training and inference).
- Gemini Nano: on-device AI that runs on the Galaxy XR or a paired smartphone (real-time evaluation, function calling).
- YOLOX-S · OSNet x0.25: object detection and tracking (NPU-accelerated).
- Galaxy XR integration: the passthrough camera and AR overlay put information right in the learner's view.
The way we combine these for the job of passing on cooking is the core of the patent we have filed.
4. Device roadmap
Here is an honest split between what is running today and what will come later.
- Now, Phase 1: full stack on Galaxy XR: a hybrid of a smartphone (the hub) and Galaxy XR (the real-time engine). Sharing the learner's view and the AR overlay are the Phase 1 goals we are building now.
- Phase 2–3: smart glasses: devices such as Galaxy Glass jinju and micro-LED AR glasses (Haean) will be added as options once they are released and pass performance testing (on the roadmap for now).
For more on the technology, see the Technology & Trust page. Security and privacy are covered in our Privacy Policy.