Research Robotics/Computer Vision Engineer

Skild AI • San Mateo, CA

13h • $250,000 - $300,000 • Onsite

About The Position

Skild AI, Inc. seeks a Research Robotics/Computer Vision Engineer in San Mateo, CA responsible for developing perceptive, intelligent, and adaptable robotic systems capable of learning and performing tasks with a focus on 3D computer vision and autonomous navigation. This includes designing perception pipelines, optimizing SLAM systems, and creating learning-based algorithms for robust robotic control in real-world environments.

Requirements

• Master’s degree (or foreign equivalent) in Computer Vision, Robotics, or a directly related discipline
• One (1) year of experience in Machine Learning or Data Science
• Experience with reconstructing 3D scenes using monocular videos, meshes, pointclouds, Neural Radiance Fields, and Gaussian Splats
• Experience with reconstructing rigid and articulated hand-held objects from videos, including inferring the time-varying hand configurations and relative poses of the objects
• Experience with generative computer vision, including diffusion models to guide reconstruction, or addressing occlusion and limited viewpoint variations in videos via data driven priors
• Experience optimizing attention-based models for perception used in autonomous navigation systems
• Experience using Neural Architectural Search (NAS) to find better perception backbone architectures with higher accuracies and lower latencies
• Experience with cloud-based training in AWS, Google cloud, or Vetex AI and optimized data loading for cloud based distributed training for deep learning workloads (e.g. Pytorch dataloader, or sharding) with hardware-in-loop

Responsibilities

• Implementing perception on robots to enable safe exploration and navigation in real world environments in collaboration with the locomotion team
• Reconstructing an entire scene in 3D using monocular images, estimating camera poses, optimizing and streamlining 3D SLAM
• Developing a set of software tools for localization of a robot using only visual inputs
• Building robust software to enable life-long mapping on a robot via optimally merged pose-graphs
• Visual servoing wrt objects detected/ tracked to control robot motion
• Researching novel techniques to detect and cater to glare during robotic mapping and navigation
• Building infrastructure and pipeline and collecting data to enable streaming of hand movements for training robot manipulation tasks such as pick and place
• Maintaining a camera and 2D lidar based navigation stack, including fixing bugs, adding new customer feature requests, and ensuring successful deployments.

What This Job Offers

• Job Type: Full-time
• Career Level: Senior