Responsibilities:
- Lead the design and development of computer vision-based AI solutions focused on detecting unsafe worker behaviors (e.g., not wearing helmets, no safety harness, unauthorized access) and identifying high-risk construction activities (e.g., working at heights, hot work, deep excavation).
- Develop end-to-end visual recognition pipelines using streaming video data from CCTV systems, body-worn cameras, drones, and other on-site sources.
- Design, train, and optimize deep learning models for object detection, human pose estimation, action recognition, and anomaly detection with high accuracy and low latency.
- Drive AI models from prototype to production deployment, including model optimization for edge computing devices and integration with existing infrastructure.
- Collaborate closely with product managers, hardware engineers, and field operations teams to understand site-specific challenges and ensure practical AI deployment.
- Build and manage data workflows, including data collection, labeling, augmentation, quality control, and versioning for large-scale training datasets.
- Stay up to date with cutting-edge research (e.g., Vision Transformers, self-supervised learning, few-shot learning) and apply innovative techniques to improve model generalization and cross-scene adaptability.
- Mentor junior AI engineers and establish best practices for efficient, scalable, and maintainable AI development.
Requirements:
- Degree or Higher Diploma in Computer Science, Artificial Intelligence, Pattern Recognition, Automation or related disciplines.
- Minimum of 5 years of professional experience in AI, with a strong focus on computer vision and deep learning.
- Proven experience in delivering AI solutions from concept to production, especially in real-world applications.
- Proficient in deep learning frameworks such as PyTorch or TensorFlow, and experienced with computer vision libraries like OpenCV, MMDetection, MMPose, YOLO, etc.
- Solid understanding of core vision tasks: object detection, pose estimation, action recognition, and video analysis.
- Hands-on experience with video stream processing (RTSP, ONVIF), multi-camera fusion, and model deployment on edge devices (e.g., NVIDIA Jetson, Huawei Atlas).
- Familiarity with model optimization techniques for low-latency and resource-constrained environments (e.g., TensorRT, OpenVINO, NCNN).
- Demonstrated ability to lead small technical teams and manage project timelines independently.
- Strong ability to read technical literature and write clear documentation in English.
- Experience in smart construction sites, industrial safety monitoring, or video surveillance systems is highly preferred.
- Prior exposure to construction environments or field testing is a strong plus.
- Experience with drone-based video analysis or multi-modal fusion (vision + sensors) is advantageous.
- Publication record in top-tier conferences (e.g., CVPR, ICCV, ECCV, NeurIPS, AAAI) is a bonus.