Develop novel formulations and architectures for a wide variety of computer vision tasks
Perform large-scale distributed training of deep neural networks to build a unified and consistent vector space for autonomous driving tasks (e.g., occupancy, occupancy flow, semantics, geometry, detection, drivable surface)
Design metrics, tasks, and datasets that aid in perception and autonomy
Deploy models at scale to millions of Tesla cars in the real world
Strict adherence to strong software engineering practices to develop novel work quickly and safely
What You’ll Bring
Strong experience writing production-level Python and software engineering best practices
Solid mathematical fundamentals including linear algebra, vector calculus, probability theory, and numeric optimization
An “under the hood” knowledge of deep learning: layer details, loss functions, optimization, etc
Understanding of modern deep learning techniques (CNNs, transformers, autoregressive models, etc.)
Domain expertise in at least one of these areas: object detection & tracking, pose estimation, depth estimation, semantic & instance segmentation, video models, differentiable rendering, Neural Radiance Field (NeRF), 3D reconstruction, visual SLAM, structure from motion
Familiarity with basic computer vision concepts such as intrinsic and extrinsic calibrations, homogeneous coordinates, projection matrices, and epipolar geometry
Experience with PyTorch, or at least another major deep learning framework such as TensorFlow
Experience in GPU programming (e.g., CUDA, OpenCL, OpenGL) or GPU-accelerated libraries
Comfortable working in a shared cluster environment