VinRobotics
AI tóm tắt
Skills bắt buộc
Job Description Develop, test, and deploy reliable AI models for automated visual inspection in automotive and manufacturing environments—covering defect detection, quality validation, and real-time decision-making under industrial constraints.
Key Responsibilities:
Develop, test, and deploy advanced computer vision and deep learning models tailored for industrial inspection tasks, including surface defect detection (scratches, dents, cracks), assembly verification, and anomaly detection.
Develop, test, and deploy robust AI solutions capable of operating reliably under real-world manufacturing conditions such as varying lighting, noise, occlusions, and product variability.
Collaborate with hardware and automation teams to define camera systems, lighting setups, sensor configurations, and data acquisition strategies for optimal inspection performance.
Collaborate with cross-functional teams to design and implement data pipelines, including dataset collection, annotation workflows, versioning, and validation frameworks with clearly defined metrics (precision, recall, false rejects/accepts).
Deliver production-grade AI systems optimized for edge deployment with strict latency, stability, and accuracy requirements.
Contribute to continuous improvement of inspection performance through failure analysis, model retraining, and feedback loops from production environments.
Required Qualifications:
BS+ in Computer Science, AI, Robotics, Electrical Engineering, or related field.
Strong Python; experience with deep learning frameworks (PyTorch preferred, or TensorFlow/JAX).
Solid computer vision foundation: image processing, object detection (YOLOv8/v9/v10/v11, RT-DETR, DETR-family), segmentation (Segment Anything Model (SAM/SAM2), U-Net, Mask R-CNN), anomaly detection techniques (PatchCore, EfficientAD, PaDiM, FastFlow).
Experience handling industrial imaging data: cameras (area/line scan), optics, lighting, calibration, and image preprocessing.
Experience building reliable ML systems: dataset curation, data augmentation, model validation, and performance optimization.
Familiarity with deploying models in production environments (Linux, Docker, REST/gRPC APIs).
Strong debugging and analytical skills for diagnosing model failures and improving robustness.
Preferred Qualifications:
Experience in automotive or manufacturing quality inspection systems.
Exposure to edge AI deployment and model optimization (NVIDIA Jetson Orin, TensorRT, ONNX Runtime, OpenVINO, model quantization/pruning, knowledge distillation).
Exposure to vision-language and foundation models for inspection (e.g. Grounding DINO, CLIP-based approaches for few-shot defect classes).
Experience with synthetic data generation, diffusion models/GANs, or simulation (e.g. NVIDIA Omniverse/Isaac Sim/Blender) for defect modeling.
Understanding of quality standards (Six Sigma, SPC, zero-defect manufacturing concepts).
Experience with 3D vision, multi-camera systems, point cloud processing, or multimodal inspection.
Familiarity with MLOps tooling for experiment tracking and model versioning (MLflow, Weights & Biases, DVC).
Tools & Stack
Python, PyTorch / TensorFlow, OpenCV
ONNX, ONNX Runtime, TensorRT, OpenVINO
Industrial cameras (Basler, Cognex, Keyence, etc.), lighting systems
Docker, Linux, REST/gRPC APIs
Git, CI/CD pipelines
MLOps: MLflow / Weights & Biases (optional)
Edge AI platforms: NVIDIA Jetson Orin, IPC (optional)
Nguồn: careers VinRobotics
