Leading development of GenP, an end-to-end Physical AI humanoid product built on a common core with domain profiles for customer-specific deployments.
Context
- GenON is building GenP, a Physical AI humanoid product delivered to customers as domain-specific, end-to-end solutions.
Problem
- Voice conversation, VLA manipulation, SLAM navigation, and whole-body control had to run together in one product through a single workflow.
- Each customer domain needs its own configuration while the core stays shared.
My role
- Product Owner (PO) and development lead, owning the product definition, commercialization roadmap, architecture, and data and model pipeline.
Approach
- Defined the product as a common core plus domain profiles, so customer-specific end-to-end products are composed from shared modules.
- Designing an architecture that integrates voice conversation, VLA manipulation, SLAM navigation, and whole-body control into a single workflow.
- Designing and building a control panel for robot registration, status monitoring, logs and alarms, task execution, and emergency stop.
- Setting up the data and model pipeline, from Teleop data collection to VLA training and evaluation and a simulation environment for validation.
Architecture
| Layer | Components | Flows |
|---|---|---|
| 01 Operations |
| to 03 Workflow · Tasks |
| 02 Domain profiles |
| to 03 Workflow |
| 03 Workflow |
| to 04 Common core |
| 04 Common core |
| to 05 Robot |
| 05 Robot |
| to 06 Data and model pipeline · Teleop data |
| 06 Data and model pipeline |
| back to 04 Common core · Trained models |
Outcome
In progress- In progress. The product definition and the phased commercialization roadmap are written; the architecture, control panel, and pipeline are under development.