Senior Applied AI / Data Scientist – Robotics & Embodied Agents 

The Opportunity 

Join our ML research and engineering team to develop the intelligence layer of ATOM. You'll design and train domain-specialized models that reason over structured engineering constraints, simulation outputs, and synthetic data. Your work directly improves agent reasoning, reliability, and the ability to solve novel robotic automation problems in high-fidelity simulated and real-world environments. 

What You'll Own 

• Develop and optimize agent architectures that integrate multimodal perception, reasoning over structured geometric and constraint data, and precise real-time execution. Leverage simulation of fidelity and synthetic data to train models that transfer to real robotic systems. 

• Design and scale training recipes using supervised fine-tuning, reinforcement learning, imitation learning, and in-context learning. Build data pipelines that leverage simulation-generated synthetic datasets and real-world demonstration data. 

• Design memory systems and planning mechanisms that enable agents to reason over long horizons, maintain state across complex multi-step automation tasks, and effectively leverage context from previous interactions. 

• Research and implement capabilities that allow agents to adapt to novel environments and learn from experience at test time, improving robustness to distribution shifts and new automation scenarios. 

• Establish rigorous evaluation protocols within NVIDIA Isaac Sim and our custom simulation environments. Design benchmarks that measure agent generalization, reasoning quality, and real-world transfer. 

Required Experience 

• Bachelor's, Master's, or Ph.D. in Computer Science, Artificial Intelligence, Machine Learning, or a closely related field 

• Strong hands-on experience with modern ML frameworks (JAX, TensorFlow, or PyTorch) 

• Strong Python programming skills with experience building and maintaining large-scale data pipelines 

• Deep understanding of LLM internals: training pipelines, computational characteristics, multimodal extensions, and fine-tuning strategies 

• Solid knowledge of Deep Reinforcement Learning, LLM reasoning, Imitation Learning, Memory-based architectures, Vision-Language Models (VLM), and/or Vision-Language-Action (VLA) models 

• Proven track record of designing and maintaining robust technical assets libraries, frameworks, or models used by large technical teams; open-source contributions are a plus 

• Excellent communication and collaboration skills 

Minimum 5 Years Professional Experience 

• Building embodied agents for 3D virtual environments, physics simulators, or robotics platforms 

  Strong track record in ML/AI competitions or peer-reviewed publications at top-tier venues (NeurIPS, ICLR, ICML, CVPR, ICRA, etc.) 

• Applied ML engineering with modern frameworks and production-scale model training 

• Strong Plus 

  • Experience designing embodied agents, evaluating in simulators, and transferring to real robots 

  • Background in robotics, motion planning, or physics simulation 

  • Experience embedding physical, geometric, or domain constraints into neural models 

  • Hands-on experience with simulators like NVIDIA Isaac Sim, PyBullet, or MuJoCo 

  • Exposure to reinforcement learning, planning systems, or hierarchical control 

  • Demonstrated passion for AI making measurable impact in real-world robotics