Responsibilities
1. Responsible for the model training standards and corpus production of ecological governance model data (video, graphics, text, etc.), building a suitable Prompt model, conducting experiments and tests, and continuously optimizing the performance and effect of the model 2. Precipitating annotation skill trees through problem summary and standard optimization to ensure that the data quality meets the requirements 3. Work closely with algorithm and product teams to convert algorithm requirements into model corpus documents that can be mass-produced, and output reasonable and accurate annotation rules and evaluation standards 4. Understand industry trends, mine high-quality data sources, evaluate the gain of each source on model performance, and support the improvement of model quality.
Qualifications
1. Have good business understanding ability, be able to understand requirements from a business perspective, understand model capability boundaries and optimization directions, and be able to translate them into technical language 2. Understand large language model technology, and have participated in the training and optimization iteration process of large language models, including pre-training, SFT, and reinforcement learning 3. Be proficient in prompt writing, application, and optimization skills, and understand Langchain, AI Agents principle, interested in human-machine assistance and action 4. Have exploratory spirit, strong self-learning ability, and continue to learn and try new things.
