Assist in building and maintaining training datasets: data collection, cleaning, deduplication, and checking annotation quality.
Run and reproduce existing training scripts: execute training under guidance, record logs, and organize training results (loss, metrics).
Implement and tune data augmentation and preprocessing pipelines (implement or adjust common augmentations).
Help integrate annotation formats (COCO, Pascal, YOLO) and evaluation scripts, and produce standardized validation reports.
Support preparation of synthetic data, data-augmentation workflows (using simple scripts or toolchains).
Under the guidance of senior engineers, perform basic model export and packaging (e.g., ONNX, TorchScript or similar simple formats).
Write or supplement experiment logs, training-step instructions, and data documentation to ensure reproducibility.
Required Qualifications
Master’s student or recent graduate; majors in Computer Science, Software, Electrical, Computer Engineering, Automation, Machine Learning, or related fields preferred.
Proficient in Python and able to write scripts for data processing (basic operations with Pandas, NumPy, OpenCV).
Basic familiarity with a deep-learning framework (PyTorch or TensorFlow); able to run training, evaluation scripts provided by others.
Understanding of common image data formats and annotation formats (COCO, Pascal VOC, YOLO).
Good habit of keeping experiment records; able to write clear reproduction steps in Markdown or plain text.
Basic Linux , command-line skills; experience using Git for version control.
