Machine Learning Engineer: Perception

About this role

Bedrock is seeking a Machine Learning Engineer to develop production-grade 3D perception systems for autonomous construction machinery. You'll design and deploy multi-modal fusion architectures (LiDAR and camera) that handle real-world challenges like occlusion, harsh weather, and vibration, while optimizing models for edge hardware.

What you'll do

• Design and train early-fusion perception models (BEV transformers) combining LiDAR and camera data for object detection and segmentation
• Build robust systems handling dynamic occlusion, particulates, and high-vibration construction environments
• Optimize and deploy deep learning models to embedded hardware with attention to latency and inference constraints
• Debug system-level issues including sensor calibration drift and real-time performance bottlenecks
• Manage data pipelines and ensure ground truth quality for model training and evaluation
• Collaborate with cross-functional teams to create perception representations for downstream autonomy tasks

What they're looking for

• Deep learning frameworks (PyTorch, TensorFlow, or JAX)
• 3D geometry and multi-sensor calibration (SE(3), homogeneous coordinates, intrinsic/extrinsic calibration)
• Early-fusion architectures (BEVFusion, TransFuser, PointPainting)
• Transformer-based object detection (DETR, PETR, temporal variants)
• Python and systems programming (C++ or Rust)
• Data pipeline management and ground truth annotation
• LiDAR and computer vision perception
• Edge deployment and real-time optimization

Benefits

• Work on meaningful real-world autonomy problems in construction
• Collaborate with veterans from Waymo, Segment, and Uber Freight
• Well-funded company ($350M backing) with resources to execute at scale
• Located in San Francisco with flexible remote considerations
• Opportunity to impact billion-dollar infrastructure projects