Machine Learning / Computer Vision Engineer

About the company Eka Robotics is focused on building intelligent robotic systems for the physical world, aiming for speed, generalization, and reliability. Their approach is grounded in physics and seeks to unlock superhuman robotic capabilities. The team consists of experts in robotics and machine learning working at the frontier of research and deployment. About the role The company is hiring a Machine Learning / Computer Vision Engineer to help scale its R&D efforts. The role involves developing vision systems and learning-based models that enable robotic manipulation and real-world deployment. It is a hands-on position requiring both research and production-level contributions. Responsibilities - Build computer vision and visual representation learning pipelines for robotic manipulation across RGB, RGB-D, depth, segmentation, pose, keypoint, and object-centric data. - Develop visual models supporting reinforcement learning and imitation learning, including end-to-end visuomotor policies. - Improve data pipelines through domain randomization, synthetic data generation, rendering, sensor noise modeling, and real-world fine-tuning. - Design robust perception models resilient to lighting, viewpoint, occlusion, clutter, and calibration issues. - Evaluate models on real robotic tasks, identify failure modes, and iterate on improvements. - Collaborate with robotics and simulation engineers to define perception strategies. - Set up and evaluate camera and depth sensing systems with focus on real-world robustness. Requirements - Ph.D. in computer vision or 3+ years of relevant industry experience. - Strong background in deep learning-based computer vision. - Experience with frameworks such as Jax or PyTorch. - Experience in visual representation learning, detection, segmentation, pose estimation, or 3D perception. - Strong Python programming skills. - Ability to transition between research and production systems. - Understanding of data distribution, sensor noise, calibration, and environmental variation. Nice to have - Experience training policies from visual inputs including RGB-D, point clouds, or latent representations. - Experience with synthetic data, rendering, and simulation tools such as Isaac Sim or MuJoCo. - Familiarity with reinforcement learning, behavior cloning, or diffusion policies. - Experience with real robot deployment, ROS/ROS2, and calibration workflows. - Publications in top-tier ML, CV, or robotics conferences.