We are a research group at Southeast University working on efficient, evolvable, and deployable AI systems.
Our research focuses on two closely connected directions:
- Embodied AI memory and self-evolution: enabling embodied agents to accumulate experience, maintain long-term memory, and improve continually through interaction.
- On-device AI infrastructure: building efficient inference, adaptation, and resource-management systems for heterogeneous robots and edge devices.
Embodied.cpp is a portable C++ inference runtime for embodied AI models, including vision-language-action models and world-action models. It provides a shared deployment path across heterogeneous CPUs, GPUs, NPUs, robots, and simulators.
- Long-term memory and continual evolution for embodied agents
- Efficient inference and adaptation on edge devices
- Portable AI runtimes for heterogeneous hardware
- Resource scheduling for AI systems
- Edge intelligence and serverless AI infrastructure
We welcome research collaboration and open-source contributions related to embodied AI systems and AI infrastructure.
To display this README on the organization homepage, place it in the special repository:
SEU-PAISys/.github/profile/README.md