Sr. Software Engineer, Perception Data Infrastructure

Nuro Nuro · Robotics · CA · Autonomy

Nuro is seeking a Sr. Software Engineer for their Perception Data Infrastructure team. This role focuses on building and owning the core native perception data platform for autonomous vehicles, bridging hardware, labeling operations, and ML models. The engineer will handle massive 3D point clouds, complex sensor parsing, and pipeline propagation with strict performance constraints, ensuring high-quality data for ML training.

What you'd actually do

  1. Design and advance systems that: Render and manipulate massive, multi-modal sensor datasets (Lidar, Camera, Radar) efficiently in UI environments.
  2. Parse raw sensor data and manage complex data pipeline propagation between onboard hardware logs and offboard visualization.
  3. Enforce strict data validation and annotation constraints to prevent upstream labeling errors from polluting downstream ML pipelines.
  4. Create seamless, resilient API boundaries between our ML data pipelines and our Ground Truth/Data Platform infrastructure.

Skills

Required

  • 4+ years of industry software engineering experience
  • Fluency in C++
  • strong familiarity with Python
  • Proven ability to lead cross-functional technical projects, bring order to complex codebases, and drive them to completion in ambiguous settings.
  • deep practical experience in building robust, high-performance software systems or infrastructure.

Nice to have

  • Deep familiarity with robotics software stacks, sensor parsing, coordinate frames, and pipeline propagation.
  • Familiarity with 3D graphics/rendering (WebGL, OpenGL), point cloud processing, or autonomous driving sensor data.

What the JD emphasized

  • massive, dense 3D point clouds
  • complex sensor parsing
  • pipeline propagation
  • rigid performance constraints
  • highly ambiguous environments
  • complex systems
  • strict API boundaries
  • good enough, fast enough
  • high-quality data
  • massive, multi-modal sensor datasets
  • complex data pipeline propagation
  • strict data validation and annotation constraints
  • upstream labeling errors
  • downstream ML pipelines
  • seamless, resilient API boundaries
  • ML data pipelines
  • Ground Truth/Data Platform infrastructure
  • Critical Path Owner
  • complex, evolving ecosystem
  • order to ambiguity
  • critical operational bottlenecks
  • sustainable long-term solutions
  • High-Velocity Impact
  • engineering lifecycle
  • autonomy progress
  • trade-off between over-engineering
  • highly performant, pragmatic integration
  • data pipelines flowing
  • autonomy’s hardest data bottlenecks
  • systems rigor
  • infrastructure excellence
  • native, C++ based perception data visualization and annotation platforms
  • deep performance bottlenecks
  • high-density Lidar point clouds
  • multi-camera projections
  • real-time
  • Defensive Engineering
  • robust data validation layers
  • sensor calibration errors
  • synchronization failures
  • human labeling mistakes
  • ML training loop
  • Cross-Functional Leadership
  • technical bridge
  • ML, Data and Infrastructure engineers
  • labeling operations
  • Language Fluency
  • Fluency in C++
  • strong familiarity with Python
  • non-negotiable
  • Ambiguity Navigator
  • lead cross-functional technical projects
  • order to complex codebases
  • drive them to completion
  • ambiguous settings
  • Systems Thinker
  • deep practical experience
  • building robust, high-performance software systems or infrastructure
  • shipping integrated, impactful solutions
  • theoretical experimentation
  • Robotics Expertise
  • Deep familiarity with robotics software stacks
  • sensor parsing
  • coordinate frames
  • pipeline propagation
  • 3D graphics/rendering
  • point cloud processing
  • autonomous driving sensor data

Other signals

  • ML models learn from high-quality data
  • prevent upstream labeling errors from polluting downstream ML pipelines
  • catch sensor calibration errors, synchronization failures, and human labeling mistakes before they enter the ML training loop