Hybrid Big Model Prompt Understanding Algorithm Engineer

Roles & Responsibilities

1. Responsible for the work related to the command intent understanding of the generative big model, build a complete command intent classification system, and improve the command understanding ability and performance of the big model as a whole
2. Responsible for the work related to the multi-round dialogue understanding of the generative big model, including but not limited to: multi-round intent judgment, multi-round rewriting, multi-round topic tracking, etc.
3. Explore prompt tuning related technologies, including but not limited to: prompt knowledge generation methods and efficient prompt knowledge embedding methods
4. Explore the construction and operation plans around the big model prompt community, and promote the construction of a big model rich prompt ecosystem
5. Explore and follow up on cutting-edge technologies, seek technological breakthroughs, and promote the improvement and breakthrough of machine capabilities in AIGC.

Qualifications

1. Familiar with cutting-edge technologies and algorithms in natural language processing related fields, and have solid basic knowledge of NLP, including knowledge in linguistics, text processing, language models, etc.
2. Proficient in Python, familiar with Linux environment development, and proficient in applying deep learning frameworks TensorFlow or PyTorch
3. Need to have the ability to generate prompt texts, and be able to design and optimize the prompts of natural language generation models
4. Need to have data processing and analysis capabilities, and be able to conduct in-depth analysis and processing of large-scale corpus data
5. Continue to follow up on cutting-edge deep learning technologies, understand cutting-edge deep learning related algorithms, and be familiar with model structures such as Transformer
6. Have the ability to analyze problem definition and solve problems, and have continuous self-drive to face challenges
7. Have experience in training and optimizing GPT models/large models, and those with reinforcement learning experience and prompt engineering work experience are preferred.