Senior ML Engineer, 4D Engine

Own the ML models that power Avala's auto-labeling pipeline — from point cloud segmentation to 4D temporal tracking. You'll ship models that run at 89% accuracy today and push toward 95%+, directly reducing the human effort required to create ground truth data for Physical AI.

What You'll Do: • Design, train, and deploy perception models for 3D point cloud and multi-sensor annotation (LiDAR, camera, radar) • Build and improve the auto-labeling pipeline that currently achieves 89% accuracy across cuboid, polygon, and segmentation tasks • Develop temporal tracking models for 4D annotation — predicting object trajectories across sequential frames • Optimize model inference for real-time performance in the browser via WebGPU compute shaders • Collaborate with the annotation operations team to identify high-impact model improvements using production QA data • Build evaluation frameworks and A/B testing infrastructure for model quality measurement

Requirements: • 5+ years of production ML experience, with at least 2 years in 3D perception, point cloud processing, or autonomous vehicle perception stacks • Strong experience with PyTorch and production model serving (SageMaker, TorchServe, or equivalent) • Deep understanding of 3D computer vision: point cloud segmentation, object detection, multi-sensor fusion • Experience with temporal models (tracking, prediction) is highly valued • Published research or shipped production systems in 3D perception

Nice to Have: • Experience at an AV company (Waymo, Cruise, Tesla, Aurora, Zoox, Nuro) • Familiarity with Gaussian Splatting or neural radiance fields • Experience with WebGPU or GPU compute for ML inference in the browser