Applied AI Postdoctoral Researcher, 3D Embodied Agents and Gaming

About us Ego is building an Infinite Game - a persistent virtual 3D world where humanlike AI agents are able to interact with players and each other to build their own relationships, communities, and games within the game. Our embodied AI agents can perceive the world in 3D, reason like a human, and write scripting code directly into the game engine. Feel free to learn more about ego on our YC launch page: The Role We're seeking an exceptional AI researcher/engineer to join our team in developing the ego game agent architecture - a groundbreaking system for autonomous gameplay in 3D environments. This role combines cutting-edge research with practical engineering to create AI agents capable of human-level reaction times (300-500ms) in complex game worlds. Working with our team and researchers from AI Singapore and NTU's Prof. Bo An's Lab, you'll help architect a hierarchical AI system that combines high-level reasoning using multimodal LLMs with fast, low-level action models. The ideal candidate brings deep expertise in computer vision, transformer architectures, and real-time AI systems, along with practical experience shipping production ML systems. Your work will focus on developing and optimizing: Real-time perception systems using state-of-the-art computer vision models Fast vision-language-action models inspired by robotics approaches Efficient model architectures that achieve human-level reaction times End-to-end autonomous gameplay across various 3D games This role represents a unique opportunity to push the boundaries of AI gaming, building on projects like Minecraft Voyager while working within our game development ecosystem. You'll collaborate closely with our engineering team to integrate these AI systems seamlessly into our game engine, while conducting novel research that advances the field of autonomous game AI. If you're passionate about combining research-grade AI with practical engineering to create autonomous agents that can truly play games like humans do, we'd love to hear from you. Key Responsibilities Develop and implement hierarchical AI architectures combining high-level reasoning and low-level action models Design and optimize real-time computer vision systems for game object detection and tracking Create and fine-tune vision-language-action models for autonomous gameplay Collaborate with AI Singapore and NTU researchers on cutting-edge AI agent architectures Implement and optimize GUI interaction models and 3D object tracking systems Contribute to data collection, model training, and benchmark development Work closely with game developers to integrate AI systems into the game engine Required Qualifications Master's or PhD in Computer Science, AI, or related field Strong programming skills in Python and experience with deep learning frameworks such as PyTorch Expertise in computer vision and transformer architectures Experience with real-time AI systems and optimization Practical knowledge of LLMs and vision-language models Background in reinforcement learning or imitation learning Familiarity with game engines and 3D environments A strong interest in video games and a desire to contribute to the future of interactive entertainment Strong communication and collaboration skills, with the ability to explain complex technical concepts to both technical and non-technical audiences Ability to self-manage and work independently or collaboratively as needed Preferred Skills (Nice to Have) Experience with vision-language-action models (VLA) Knowledge of model distillation and optimization techniques Familiarity with YOLO, SAM, or similar computer vision frameworks Experience with behavior cloning and inverse dynamics models Background in game development or 3D graphics Publication record in relevant conferences (ICLR, NeurIPS, ICML, etc.) Project Highlights You'll be working on: Developing real-time AI agent architectures with 300-500ms latency Implementing multi-modal LLM systems for game understanding Creating efficient vision-language-action models Building scalable data collection and training pipelines Benchmarking across various 3D games and environments #J-18808-Ljbffr ego AI (YC W24)