Senior Software Engineer, Perception (Robotics)

Grab; Singapore; On-site; SGD 120,000 - 180,000; Full time

Job summary

Grab is seeking a seasoned researcher/engineer in perception and prediction for autonomous driving. You will develop and improve multi-object tracking, sensor fusion, and world models for real-world operation.

Collaborate with cross-functional teams to productionize perception and prediction capabilities across camera, LiDAR, radar, and IMU systems, focusing on accuracy, robustness, and low latency in a Singapore-based setting.

Qualifications

• 3+ years shipping perception or prediction systems for robotics, AV, ADAS.
• Strong experience with classical and learning-based perception approaches and hybrid design.
• Proven depth in multi-object tracking: data association, Bayesian filtering, ID stability.
• Hands-on experience with generic perception, occupancy representations, and unknown-object handling.
• Hands-on depth in motion forecasting, occupancy flow, streaming BEV, or world models.
• Fundamentals in geometry, multi-sensor systems, and diagnostic calibration skills.
• Proficiency in C++ and Python for production-quality embedded code.
• Experience with reproducible experiments, metrics-driven analysis, and validation-focused execution.
• Self-motivated with genuine enthusiasm for autonomous driving and embodied AI.
• Proficiency leveraging AI tools to accelerate engineering without compromising quality.

Responsibilities

• Develop and improve multi-object tracking using classical and learning-based approaches.
• Build generic world understanding for class-agnostic obstacle detection and occupancy representations.
• Build motion prediction for multi-agent trajectory forecasting with uncertainty quantification.
• Develop and evaluate end-to-end temporal perception-prediction models with modular baselines.
• Explore and adapt VLMs and foundation models for open-vocabulary robot perception.
• Strengthen multi-sensor and temporal integration across camera, LiDAR, radar, and IMU.
• Define quality gates and drive down regressions using fleet-log root-cause analysis.
• Partner with cross-functional teams to productionize perception and prediction capabilities.

Skills

Perception systems

Multi-object tracking

C++

Python

Sensor fusion

Validation & metrics

Tools

OpenCV

PCL

ROS 2

NVIDIA Orin

TensorRT

Job description

What you'd do:

• Develop and improve multi-object tracking using classical and learning-based approaches.
• Build generic world understanding for class-agnostic obstacle detection and occupancy representations.
• Build motion prediction for multi-agent trajectory forecasting with uncertainty quantification.
• Develop and evaluate end-to-end temporal perception-prediction models with modular baselines.
• Explore and adapt VLMs and foundation models for open-vocabulary robot perception.
• Strengthen multi-sensor and temporal integration across camera, LiDAR, radar, and IMU.
• Define quality gates and drive down regressions using fleet-log root-cause analysis.
• Partner with cross-functional teams to productionize perception and prediction capabilities.

What they want:

• 3+ years shipping perception or prediction systems for robotics, AV, ADAS.
• Strong experience with classical and learning-based perception approaches and hybrid design.
• Proven depth in multi-object tracking: data association, Bayesian filtering, ID stability.
• Hands-on experience with generic perception, occupancy representations, and unknown-object handling.
• Hands-on depth in motion forecasting, occupancy flow, streaming BEV, or world models.
• Fundamentals in geometry, multi-sensor systems, and diagnostic calibration skills.
• Proficiency in C++ and Python for production-quality embedded code.
• Experience with reproducible experiments, metrics-driven analysis, and validation-focused execution.
• Self-motivated with genuine enthusiasm for autonomous driving and embodied AI.
• Proficiency leveraging AI tools to accelerate engineering without compromising quality.

Nice to have:

• End-to-end joint detection-tracking-forecasting, occupancy prediction, and streaming BEV representations.
• Motion prediction expertise: multi-modal trajectories, interaction-aware forecasting, uncertainty calibration.
• Experience with VLMs, VLAs, embodied foundation models, and open-vocabulary perception.
• Foundation model adaptation: fine-tuning, distillation, quantisation, and edge deployment.
• 3D/BEV detection, sensor fusion (BEVFusion, LSS), and temporal models.
• Experience with OpenCV, PCL, ROS 2, NVIDIA Orin, TensorRT platforms.
• Familiarity with model lifecycle: dataset curation, auto-labeling, regression detection.

Additional requirements (duplicate section):

• 3+ years shipping perception or prediction systems for robotics, AV, ADAS.
• Strong experience with classical and learning-based perception approaches and hybrid design.
• Proven depth in multi-object tracking: data association, Bayesian filtering, ID stability.
• Hands-on experience with generic perception, occupancy representations, and unknown-object handling.
• Hands-on depth in motion forecasting, occupancy flow, streaming BEV, or world models.
• Fundamentals in geometry, multi-sensor systems, and diagnostic calibration skills.
• Proficiency in C++ and Python for production-quality embedded code.
• Experience with reproducible experiments, metrics-driven analysis, and validation-focused execution.
• Self-motivated with genuine enthusiasm for autonomous driving and embodied AI.
• Proficiency leveraging AI tools to accelerate engineering without compromising quality.