Computer Vision Engineer

As a Computer Vision Engineer, you will own the perception stack: turning noisy raw video, IMU, and gaze data into structured, labeled signals that are useful for training robotics and embodied AI models.

• Develop vision pipelines for: i) Human pose estimation, gaze tracking, and hand-object interactions, ii) Activity segmentation from egocentric video, iii) Optical flow, SLAM, and depth estimation (where applicable)
• Implement or adapt state-of-the-art models (e.g., YOLOv8, Detectron2, MediaPipe, DeepLabCut, SlowFast, etc.)
• Work with multi-modal data (video + IMU + gaze)
• Help design data collection protocols (what to record, when, how)
• Evaluate perception accuracy and performance on in-the-wild data
• Optimize pipelines for on-device or edge processing if needed

• Strong experience in Computer Vision and deep learning (2+ years)
• Proficient in Python, OpenCV, PyTorch, and common vision frameworks (e.g., MMDetection, MediaPipe)
• Hands-on with real-world video processing (noise, jitter, occlusion)
• Deep knowledge of pose estimation, object detection, segmentation
• Experience working with egocentric or wearable camera data is a huge plus
• Familiarity with multi-modal sensor fusion (IMU + video)
• Ability to work in an early-stage, fast-moving team and take ownership

• Experience with embodied AI, robotics, or imitation learning pipelines
• Worked with datasets like Ego4D, EPIC-KITCHENS, RoboNet, or BEHAVIOR
• Background in SLAM, depth estimation, or 3D reconstruction
• Experience deploying vision models on embedded devices (Jetson, Snapdragon, etc.)
• Contributions to open-source CV libraries or research papers