VinDynamics Tech Stack
Humanoid robots designed for home assistance, security, and service applications
VinDynamics is building humanoid robots backed by Vingroup, Vietnam's largest technology conglomerate, with leadership experienced across academia and industry robotics. The tech stack reveals a simulation-first R&D operation: MuJoCo, PyBullet, Gazebo for physics simulation paired with PyTorch, TensorFlow, JAX for learning, plus React/TypeScript for robot interfaces. Active projects center on sim-to-real transfer, reinforcement learning for locomotion, and LLM-based reasoning — indicating a focus on closing the gap between lab training and real-world deployment. The hiring mix skews heavily senior (5 of 8 roles) across engineering and research, with concurrent work on data systems and marketplace infrastructure.
Tech Stack 69 technologies
What VinDynamics Is Building
◆Challenges
- Sim-to-real transfer challenges
- Developing autonomous decision-making
- Robustness domain randomization
- System identification challenges
▲Active Projects
- Sim-to-real transfer
- High-fidelity simulation environment design
- Locomotion reinforcement learning algorithms
- Humanoid locomotion rl algorithms
- Design agentic cognitive architectures for humanoid robots
- Build llm-based reasoning and planning loops
- Develop robot memory systems
- Data collecting system
- Data marketplace
- Merchant onboarding & management
