Model Data Researcher

Job Description

1. Responsible for the collection, cleaning, labeling, and formatting of multimodal data (text, speech, images, video, etc.). Build efficient data processing workflows to support model training and inference;

2. Call upon and research cutting-edge multimodal models/large language models (such as chatGPT, SD, etc.), and understand the model's capability boundaries;

3. Provide high-quality data support for multimodal model training and optimize model performance;

4. Design prompts to generate target data and optimize result quality.

Job Requirements

1. Familiar with multimodal data cleaning, labeling, and loading, and understand data optimization techniques (such as TFRecord, Sharding, etc.);

2. Experience in calling large language models/multimodal models, understanding their capability boundaries and applicable scenarios;

3. Ability to design and optimize prompts to improve the quality and efficiency of generated data;

4. Familiar with data processing tools (such as Pandas, NumPy) and able to complete a full data processing workflow;

5. High standards for data quality, meticulous and responsible, able to identify and resolve data problems;

6. Familiar with Python programming, and understand Linux environment and common development tools.

Bonus points:

1. Familiar with large-scale model training and have a deep understanding of the role of data in training;

2. Have practical experience in developing multimodal/large language models;

3. Understand distributed data processing technologies (such as Spark, Flink);

4. Familiar with generative AI technologies and data annotation tools.