大模型算法（RAG/Agent)

该职位来源于猎聘 Roles and Responsibilities:

• Engage business leaders to discuss the business benefits that ML and AI can provide. Develop AI and ML models and algorithms to solve business problems
• Create and manage use cases to apply ML and AI concepts
• Monitor document ingestion pipeline, involve business to do trouble shooting
• Triage different doc ops process to make sure the on-going solution accuracy
• Address negative feedback cases and provide tech solution to mitigate
• Chunking tech fine tune if need to handle new structure of document 7 LLM fine tunning: RAG fine tune from the user feedback according to the existing query scope, execute the fine tuning (for example, enhance the intent classifier)
• Solution performance monitoring, alert the performance downgrade and provide improvement suggestion on LLM design, execute the fine tuning
• Fine tune the model, based on the continuous new golden test cases from business and pass to QA team to add into test scope
• Document best practices and solution frameworks. Replicate and scale solutions to markets with common/similar business needs

Job Requirements: Education & Experience

• Advanced degree in Statistics, Economics, Analytics, Mathematics, Computer Science, Information Technology or related field
• Solid 3 - 5 year’s professional experience building predictive and descriptive models. Global company/project work experience is preferred.
• Bilingual in Chinese and English (high fluency). Cantonese is added advantage.
• Knowledge, experience, and expertise in diverse statistical and data mining techniques 61 AI related project experience is required.
• NLP is required, especially in Large language model
• Able to finetune LLM / RAG
• Extensive experience working with large datasets - structured and unstructured
• Ability to program and review code in Python is highly desirable
• Insurance business experience and finance experience is added advantage.
• Excellent communication skill and ability to work collaboratively in a dynamic environment