Job Summary:

The VLA/VLM Engineer will design and deploy vision-language-action models that enable humanoid robots to perceive scenes, understand context, and execute actions intelligently. This role bridges computer vision, natural language processing, and decision-making — empowering robots to interpret human instructions, recognize behaviors, and perform real-world tasks safely and autonomously.

Position: Vision-Language-Action (VLA/VLM) Engineer
Department: AI application
Location: Gia Lam, Ha Noi
Reports to: Head of AI Application.

Key Responsibilities:

• Design, train, and deploy vision-language-action models (VLA) that combine visual perception, language understanding, and robotic control.
• Develop multimodal AI pipelines that process RGB/Depth video, LiDAR, and audio data for perception and reasoning
• Implement VLMs (e.g., LLaVA, GPT-4V, BLIP-2, Florence-2, Kosmos-2, or Gemini-based models) for scene understanding and natural interaction with humans.
• Integrate AI models with the humanoid robot’s control and decision layers, enabling the robot to interpret human commands and respond intelligently.
• Build and optimize prompt-based reasoning and visual grounding systems for human-robot dialogue and situational awareness.
• Conduct experiments in simulation and real-world environments to test perception–reasoning–action loops.
• Collaborate with cross-functional teams (Vision, RL, Motion, Control, Conversation) to ensure seamless end-to-end integration.

Requirements:

Must Have:

• Bachelor’s, Master’s, or PhD degree in Computer Science, Artificial Intelligence, Robotics, or a related field.
• Strong programming skills in Python and C++; experience with PyTorch or TensorFlow.
• Solid understanding of multimodal learning (vision + language + action) and transformer architectures (ViT, CLIP, BLIP, Flamingo, or LLaVA).
• Hands-on experience training or fine-tuning VLMs or LLMs with vision input for image/video captioning, grounding, or reasoning.
• Experience with dataset preparation and annotation for multimodal tasks (e.g., VQA, instruction-following, embodied navigation).
• Knowledge of deployment and inference optimization on edge hardware such as NVIDIA Jetson, AGX Orin, or RTX platforms.
• Familiarity with ROS/ROS2, real-time inference, and integrating AI perception modules into robotic systems.
  
  

Nice to Have:

• Experience with Vision-Language-Action models (RT-2, OpenVLA, PaLM-E, or GR-1).
• Research or publications in robotics, multimodal AI, or embodied intelligence (CVPR, ICRA, NeurIPS, CoRL, RSS).
• Experience in robotic control via language (e.g., natural language navigation, human instruction following).
• Understanding of reinforcement learning with multimodal feedback.
• Familiarity with safety alignment and hallucination mitigation for large vision-language models.

Benefits:

• Competitive salary and benefits package (open to salary negotiations).
• Opportunities for professional development and career growth.
• Flexible work arrangements.
• A collaborative and innovative work environment where ideas are valued and creativity is encouraged.