Perception Engineer

ENGGSOL PTE. LTD.; Singapore; On-site; SGD 90,000 - 130,000; Full time

Job summary

ENGGSOL PTE. LTD. in Singapore seeks a candidate to advance computer-vision capabilities for edge deployment in complex sensor environments. You will develop detection, tracking, and multimodal analytics, and optimize models for ONNX, TensorRT/CUDA, and Jetson platforms.

The role requires solid Python and ML/vision background, with optionally C++ skills, and hands-on experience with PyTorch, OpenCV, and model evaluation. Collaboration with engineering teams and field trials is expected.

Qualifications

• Degree in CS/AI/Engineering/Robotics or related field.
• Strong Python and machine-learning / computer-vision background.
• C++ capability would be advantageous.
• Hands-on experience in object detection, tracking, image/video analytics.
• Familiarity with dataset handling, evaluation, and metrics.
• Experience with PyTorch, OpenCV, ONNX, TensorRT / CUDA is beneficial.

Responsibilities

• Develop and improve computer-vision capabilities: object detection, classification, and multi-object tracking.
• Integrate perception outputs with AIS, radar, and system metadata.
• Create datasets, annotation practices, and evaluation sets for field conditions.
• Analyze false positives/negatives, dataset shift, and failure modes to drive improvements.
• Develop multimodal AI, behaviour analytics, and operator-support features with clear value.
• Optimize and package models for edge deployment using ONNX, TensorRT/CUDA on Jetson.
• Support experiments with vision-language models, synthetic data, and active learning.

Skills

Python

Machine learning

Computer vision

C++

Education

Bachelor's degree in CS/AI/Engineering/Robotics

Tools

PyTorch

OpenCV

ONNX

TensorRT/CUDA

NVIDIA Jetson

Job description

Key Responsibilities

• Develop and improve computer-vision capabilities including object detection, classification, multi-object tracking and false-alert reduction.
  
• Work with EO/IR and thermal imagery and, where useful, combine perception outputs with AIS, radar tracks and other system metadata.
  
• Build representative datasets, annotation practices, evaluation sets and repeatable performance benchmarks for field conditions.
  
• Analyse false positives, false negatives, dataset shift and operational failure modes, and translate findings into measurable improvements.
  
• Develop and evaluate multimodal AI, behaviour/event analytics and operator-assist functions where they provide clear operational value.
  
• Optimise and package models for practical edge deployment using ONNX, TensorRT / CUDA or equivalent, working with the Robotics Software Engineer.
  
• Support controlled experimentation with emerging techniques such as vision-language models, synthetic data and active learning, as well as integration tests and field trials.
  

Requirements

• Degree in Computer Science, Artificial Intelligence, Computer / Electrical Engineering, Robotics or a related discipline.
  
• Strong foundation in Python and practical machine-learning / computer-vision development; C++ capability would be advantageous.
  
• Hands‑on experience in one or more of: object detection, tracking, image/video analytics, thermal imaging, multimodal AI or sensor fusion.
  
• Comfortable working with datasets, model evaluation, error analysis and quantitative performance metrics.
  
• Experience with PyTorch or equivalent, OpenCV, ONNX, TensorRT / CUDA, NVIDIA Jetson or edge deployment would be advantageous.
  
• Candidates with strong project, internship or practical engineering experience may also be considered; applied capability matters more than years of experience.