APPLIED SCIENTIST II Jun 2021 - Present • Researching large language models (LLMs) regarding (1) novel reward modeling and preference optimization strategies for multilingual LLMs and (2) enabling LLMs to use tools and automate routines by invoking APIs • Developed dynamic data selection strategy for multilingual natural language understanding, reducing data overhead by 95% and increasing ease of model expansion to support new international markets (e.g., KO,TR,NL)•拥有6个市场和4种语言的对话代理和NLP模型的18个生产释放周期(ES,PT,JP,AR),包括数据摄入和处理以及开发以及模型开发,培训,培训,评估,部署,监视,监控,维护和修补•通过型号的级别播放的校准•逐渐启动的校准
主要关键词