Applied Computer Vision & ML Engineer

PERCEPT SOLUTIONS PTE. LTD.; Singapore; On-site; SGD 90,000 - 150,000; Full time

Job summary

PERCEPT SOLUTIONS PTE. LTD. is seeking a hands-on Computer Vision Engineer to train, fine-tune and evaluate CV models in production. You will prepare and maintain datasets, run experiments, and compile evaluation results and artefacts to drive quality improvements.

You will investigate false detections, missed events and tracking errors, identify edge cases, review labeling quality, and assess how factors like layout, lighting, and camera angles impact performance.

Qualifications

• Practical Python, deep-learning and Computer Vision experience.
• Experience with CV workflows (detection, classification, tracking, pose estimation or video analytics).
• Understanding of dataset preparation, model training, precision, recall, F1 and latency.
• Ability to analyze model failures and propose improvements.
• Familiarity with annotation tools and labeling quality checks.

Responsibilities

• Train, fine-tune and evaluate Computer Vision models.
• Prepare and maintain datasets, experiments, and model artefacts.
• Investigate false detections, misses, tracking errors and degradation.
• Identify edge cases and review annotation quality.
• Analyze how layout, lighting and camera angle affect performance.
• Support model packaging, deployment, and rollout activities.
• Perform retraining, threshold tuning and regression testing.
• Maintain model documentation and evaluation reports.
• Monitor production model performance and governance requirements.

Skills

Python

Deep learning

Computer Vision

Data analysis

Annotation quality checks

Tools

PyTorch

TensorFlow

OpenCV

Docker

ONNX

TensorRT

Job description

PERCEPT SOLUTIONS PTE. LTD. is seeking a hands-on Computer Vision Engineer to train, fine-tune and evaluate CV models in production. You will prepare and maintain datasets, run experiments, and compile evaluation results and artefacts to drive quality improvements.

You will investigate false detections, missed events and tracking errors, identify edge cases, review labeling quality, and assess how factors like layout, lighting, and camera angles impact performance.