Computer Vision Researcher / Engineer

We are seeking a driven Computer Vision Researcher/Engineer to build and optimize advanced 2D and 3D action recognition systems. In this role, you will be heavily involved in the end-to-end machine learning lifecycle—from generating and curating custom datasets to designing, training, and fine-tuning state-of-the-art models. Because this role straddles both engineering and research, you will also play a crucial part in documenting our findings and co-authoring academic papers.

Core Responsibilities

• Model Engineering: Design, implement, test, and fine-tune deep learning models specifically for 2D and 3D human action recognition (utilizing RGB video sequences, depth maps, or skeletal key point data).
• Data & Annotation Pipelines: Drive the data generation process. Manage and scale data annotation workflows using both our self-developed internal applications and online AI-assisted annotation tools.
• Research & Publication: Collaborate with the core research team to analyze experimental results, benchmark against state-of-the-art methods, and actively contribute to writing and publishing papers for top-tier computer vision conferences or journals.
• Testing & Evaluation: Rigorously evaluate model performance, debug edge cases, and optimize architectures to ensure robust performance across different environmental conditions.

Required Qualifications

Domain Expertise: Good understanding of action recognition / human activity recognition (HAR), spatial-temporal networks (e.g., 3D CNNs, Video Transformers), or Graph Convolutional Networks (GCNs) for skeleton-based tracking.

Technical Stack: Deep proficiency in Python and modern deep learning frameworks (PyTorch or TensorFlow), alongside standard vision libraries (OpenCV, MediaPipe, etc.).

Data Handling: Hands-on experience building custom datasets, handling video/3D data preprocessing, and ensuring high-quality annotation standards.

Research Acumen: Ability to read, quickly understand, and implement algorithms from recent research papers.

EA License No.: 96C4864

Reg No.: R25128798 HUANG QIMENG