Michael N.

Background

About Michael N.

I am an AI research scientist with over 20 years of cross-disciplinary R&D experience in speech recognition, natural language processing (NLP), and large-scale language models (LLMs). I began my career at Peking University with a degree in Information Science. I then pursued a Master's degree in Computer Science at the University of Southern California (USC), focusing on speech recognition and statistical machine learning. I also worked on early speech modeling projects at SRI International, participating in the development of speech recognition prototypes for human-computer interaction systems. I joined IBM's New York Lab in 2006 and have held positions as a data mining engineer, machine learning engineer, and research engineer, focusing on applying NLP and ML systems to industry-wide scenarios such as intelligent customer service, semantic search, and financial risk management. During this time, I led multiple end-to-end AI projects, becoming proficient in model development and deployment using platforms such as TensorFlow, PyTorch, and Spark. I also facilitated the patenting and transformation of numerous research findings into enterprise solutions. Since 2021, I have joined Anthropic, where I have been involved in the design of key modules for large-scale language models (the Claude series), focusing on alignment, security, and conversational capabilities, particularly within subsystems such as RLHF, Constitutional AI, and long-context understanding. I built an experimental framework for the RLHF reward model, designed a multi-dimensional user preference scoring system, and assisted enterprise clients (in finance, law, education, and other sectors) with private model deployment and fine-tuning. During this work, I collaborated closely with multiple teams, including product, security, and legal, to ensure model compliance, security, and controllability. I also actively participated in research initiatives on AI ethics and system security, publishing relevant results at top international conferences such as NeurIPS and ICLR, and was invited to present at the Stanford AI Symposium. I believe that AI is not only an engineering challenge but also a social responsibility. The boundaries of technology must be stretched in tandem with human values. I welcome discussions and collaborations with you.