Tencent Video NLP Algorithm Engineer

Roles & Responsibilities

1 .Responsible for Tencent video text content understanding tasks, including but not limited to algorithm development/implementation/industry influence of large language models in the field of content understanding
2. Combined with the needs of self-developed products, deeply participate in the research and development of large language models, Optimize the generation effect of large language models from the aspects of data set construction, model pre-training, prompt optimization, SFT, RLHF, CoT, knowledge/model editing, etc. to achieve high-quality large language models
3. Follow up on LLM generation The latest developments and applications of models, responsible for exploring and tracking the research and implementation of large language models in long text content understanding, and optimizing the effects of large language models in multiple long text content understanding businesses
4. Responsible for combining business needs Carry out data insights, algorithm research, effect evaluation/optimization, algorithm innovation, etc.

Qualifications

1. Strong algorithm implementation capabilities, familiar with deep learning platforms such as tensorflow, pytorch, etc.
2. Experience in LLM, NLP, reinforcement learning, machine learning, etc. Have a certain research foundation, be familiar with mainstream models and algorithms, pay attention to the latest developments in the field, be able to independently carry out research and development work in one or more fields, and be able to follow up and implement new algorithm capabilities
3. Have large-scale applications Practical experience in solving relevant practical problems with neural networks, able to select appropriate parameters adjustment, training task selection, training data enhancement, objective function changes and other strategies for appropriate data and model expression to ensure the success rate of training for very large models, etc., and can Explain the source of intuition behind the strategy and use experimental phenomena to verify the hypothesis
4. Priority will be given to those with strong learning competition experience and high-quality paper publications (such as NeurlPS, ACL, EMNLP, ICML, ICLR, etc.) < br/>5. Be passionate, eager to learn, and have good teamwork and communication skills.