Lead Data Scientist (LLM post training) / 资深算法科学家 （大模型后训练方向）
- Beijing
- Permanent
- Full-time
- 大模型的后训练与优化，包括微调、低资源训练、蒸馏、性能评估等；
- 领域数据的收集、清洗、标注及增强；
- 探索后训练在推理模型、多模态、多语言、智能体等领域的技术研发；
- 在面向用户业务场景和内部流程效率提升场景下的应用落地，包括风控、客服、个性化、等场景；
- 以文档、专利、原型、产品上线等多种方式呈现和展示技术成果；
- 积极参与业务和产品规划。
- LLM post-training and optimization, including fine-tuning, low-resource training, distillation, and performance evaluation.
- Domain-specific data collection, generation, cleaning, labeling, and augmentation.
- Develop technology and systems to post-train reasoning models, multimodal models, multi-lingual models and agents.
- Application deployment in user-facing product experience and internal productivity scenarios, such as risk control, customer support, personalization and others.
- Present and demonstrate technical achievements through documents, patents, prototypes, and product launches.
- Actively participating in business and product planning.
- 在计算机、人工智能相关领域取得硕士以上学历；
- 具备扎实的机器学习、深度学习理论基础，对大模型以及相关技术领域有深入理解；
- 具备大模型后训练、微调的实际经验，熟练掌握并灵活应用SFT、DPO等技术；
- 具备清洗、标注、生成处理大量训练数据的经验；
- 具备基于大模型的应用开发部署上线的经验；
- 熟练掌握Python语言，熟练使用主流深度学习框架；
- 优秀的书面和口头沟通能力，并具备良好的英语沟通能力；
- 对技术产品化和商业化有浓厚的兴趣。
- 开源社区参与和贡献经验；
- 顶级会议上论文发表经历；
- 端到端应用落地经验；
- 良好的技术大局观和业务理解能力
- Master's degree or higher in computer science, artificial intelligence, or a related field.
- Solid theoretical foundation in machine learning and deep learning, with in-depth understanding of large models and related technical domains.
- Experience in large model post-training and fine-tuning, proficient in applying techniques such as SFT and DPO.
- Experience in cleaning, labeling, and generating large amounts of training data.
- Experience in developing, deploying, and launching applications based on large models.
- Proficiency in Python and mainstream deep learning frameworks.
- Excellent written and verbal communication skills, and good English communication skills.
- Experience in participation and contribution to open-source communities;
- Publication experience at top conferences.
- Experience in end-to-end LLM application deployment.
- Good technical strategic sense and strong business acumen.