Senior Perception Engineer

John Deere John Deere · Industrial · Santa Clara, CA +1 · Product Engineering (CA)

Senior Perception Engineer for John Deere's autonomy stack, focusing on developing, maintaining, and deploying applied ML models for perception tasks (detection, segmentation, tracking, scene understanding) on embedded and edge platforms. The role involves full-stack development from sensor interfaces to ML models, diagnosing failures, optimizing C++ modules for real-time performance, and collaborating with other engineering teams. Emphasis is on production reliability and code quality.

What you'd actually do

  1. Develop and maintain perception capabilities across the full autonomy stack — working from onboard C++ systems and sensor interfaces up through applied ML models that interpret the environment
  2. Develop and maintain key components of the perception pipeline, including data ingestion, labeling, model execution, and downstream interfaces used by planning and controls
  3. Diagnose and resolve perception failures using logs and sensor data — reconstructing system behavior when real‑time debugging isn’t possible and driving issues to root cause
  4. Implement and optimize C++ modules running on embedded and edge compute platforms, ensuring real‑time performance, robustness, and clean integration with the broader autonomy system
  5. Train, evaluate, and deploy applied ML models for detection, segmentation, tracking, and scene understanding — focusing on production reliability rather than research novelty

Skills

Required

  • modern C++
  • applied machine learning for perception
  • training, evaluating, and deploying models using frameworks such as PyTorch or TensorFlow
  • working across system boundaries
  • delivering high-quality, maintainable code
  • collaborating effectively within a multidisciplinary engineering team
  • strong debugging skills, especially in log-driven environments
  • experience working with sensor data (camera, LiDAR, radar)
  • building or maintaining perception data pipelines

Nice to have

  • ROS 2
  • deploying ML models to embedded platforms
  • optimizing for latency, memory, or power constraints
  • classical computer vision
  • sensor fusion
  • tracking algorithms
  • static analysis and code quality tools
  • simulation environments
  • synthetic data generation
  • system-level testing frameworks
  • Linux development
  • build systems (CMake, Bazel, or colcon)
  • containerized workflows (Docker)
  • concurrency, performance constraints, and real-time considerations in robotics or embedded systems
  • domain experience in agricultural, mining, construction, or other off-highway autonomous systems

What the JD emphasized

  • applied ML models
  • production reliability
  • real-time performance
  • embedded and edge compute platforms
  • modern C++
  • applied machine learning for perception
  • training, evaluating, and deploying models
  • sensor data
  • perception data pipelines

Other signals

  • applied ML models
  • production reliability
  • real-time performance
  • embedded and edge compute platforms