COMPUTER VISION ENGINEER – DEVICE GRADING

We are looking for an experienced Computer Vision Engineer to develop and deploy automated device grading solutions using computer vision and deep learning techniques.

The role involves detecting and evaluating cosmetic defects such as scratches, dents, cracks, and screen damage from customer-captured images and videos under varying lighting conditions.

The ideal candidate should have strong hands-on experience with YOLO-based object detection, Python, PyTorch, model training, optimization, and production deployment.

Key Responsibilities

1. Detection Modelling

Train, fine-tune, and deploy YOLO-family object detection models.

Develop solutions for detecting small objects and low-contrast defects, including hairline cracks and scratches.

Differentiate actual surface damage from glare, reflections, and lighting variations.

Evaluate YOLO models against alternative approaches such as DETR, segmentation, and anomaly detection.

Develop defect severity scoring and classification models.

2. Data & Annotation

Manage dataset collection, curation, class balancing, and annotation strategies.

Define annotation guidelines and ensure consistent labelling quality.

Develop image augmentation techniques to handle lighting variations, reflections, motion blur, and different device finishes.

Maintain version-controlled datasets and reliable evaluation datasets.

3. Production Deployment

Deploy computer vision models into production environments.

Optimize models using ONNX and TensorRT.

Work on inference performance, latency, throughput, and model versioning.

Monitor model performance, confidence levels, and data drift.

Collaborate with backend engineers to integrate models into application workflows.

4. Model Improvement & Feedback

Develop feedback loops to capture production predictions and corrections.

Identify low-confidence predictions and model errors for improvement.

Implement continuous model retraining, evaluation, and deployment.

Prepare technical documentation and model evaluation reports.

Mandatory Skills & Requirements

4–5 years of professional experience in Computer Vision.

Strong hands-on experience training, fine-tuning, and deploying YOLO object detection models.

Strong knowledge of Python and PyTorch.

Experience with small-object detection and low-contrast image analysis.

Expertise in image augmentation and handling real-world image conditions such as glare, reflections, lighting variations, and motion blur.

Experience with production model deployment, optimization, monitoring, and versioning.

Knowledge of model evaluation metrics, confidence thresholds, and performance optimization.

Experience building feedback loops for continuous model improvement.

Ability to independently manage the complete computer vision pipeline, from raw images to production deployment.

Good to Have

Experience in device grading, visual quality inspection, or industrial defect detection.

Knowledge of anomaly detection and few-shot learning techniques such as PatchCore.

Experience with image segmentation techniques.

Exposure to video processing and multi-frame assessment.

Experience with edge or on-device inference.

Knowledge of ONNX, TensorRT, Triton, or SageMaker.

Familiarity with Vision-Language Models.

Experience with MLflow, Weights & Biases, or DVC.

Knowledge of active learning and semi-supervised learning.