About the Role
We are looking for an experienced AI Vision Consultant to design, build, and deploy a real-time hardware defect detection system on our production line. This is a hands-on technical role where you will work closely with our internal team — right from dataset strategy through to edge deployment and handover.
Key Responsibilities
- Advise on camera placement, lighting setup, and image collection protocols across defect types
- Guide data labeling and annotation using tools like Roboflow or CVAT, and design augmentation pipelines to handle class imbalance
- Train and fine-tune YOLOv8 models on hardware defect datasets and implement anomaly detection for limited-sample defect classes
- Build real-time detection pipelines using OpenCV + YOLO with accept/reject decision logic and inspection result logging
- Deploy trained models on NVIDIA Jetson Nano/Xavier and optimize for edge inference using TensorRT or ONNX
- Deliver full documentation and conduct a team handover/training session
Required Skills & Experience
- Computer Vision (object detection / defect detection) — 4+ years
- YOLOv8 / YOLO family (training, fine-tuning, export) — 2+ years
- Python with PyTorch and OpenCV — 4+ years
- Edge AI / NVIDIA Jetson deployment — 1+ year
- Data annotation (Roboflow, CVAT, or LabelImg) — 2+ years
- Model optimization (TensorRT, ONNX, quantization) — 1+ year
Good to Have
- Experience with anomaly detection methods (PatchCore, FastFlow)
- Background in industrial or manufacturing QC environments
Key Deliverables
- Camera & lighting setup guide
- Annotated dataset with augmentation pipeline
- Trained YOLO model (90%+ accuracy on validation set)
- Real-time inference pipeline on edge device
- Optimized production-ready model with full documentation
- Team training session and ongoing support
