Xian Li

About Xian Li

Lead a machine learning science team building GenAI solution for Amazon Rufus, as well as fusing customer behavior signals with LLM to enable strong personalization and recommendation experiences. Drive multiple cross-org collaboration and adoption use cases. Drive the scientific vision and deliver innovative, pragmatic solutions to intrinsically ambiguous problems. Founded and scaled three AI science initiatives from scratch into high-impact, multi-year programs. Lead execution and deliverables for organizations of 50+ people while mentoring and growing the careers of scientists and engineers across multiple teams.Expertise in LLM, NLP, Knowledge Graph with 15 years research and development experience. Published 30+ papers in top academic venues and filed 6 US patents. Serve as PC committees for top GenAI/NLP conferences and journals.I am a strategic problem solver, known for turning deep analytical insights into business impact. Combine rigorous evaluation of data quality, scalability, and experimentation with sound judgment to guide organizational strategy and execution. Recognized for navigating complexity and enabling evidence-based decisions across large cross-functional teams.

Experience

1. Principal Applied Scientist Senior Manager, Genai for Shopping

   Amazon

   Apr 2023 — Present · Seattle, WA, US

   Rufus Foundation Model: Found and lead the Rufus LLM base model training and training data curation team. Build three generations of foundation model from scratch and successfully launched it for Amazon Rufus shopping assistant. Evolve model architecture and scale up training data curation to trillions of tokens with high quality bar. Lead the team to develop scaling law based experimental toolkit to guide the organization to make data driven decisions in large scale model development. Incubate and lead the efforts of extending Amazon proprietary LLM to learn from customer behavior, develop RL-based methods to leverage customer behavior signals to enable strong personalization and recommendation experiences. Drive multiple cross-org collaboration and adoption use cases. Launched the models with multiple Amazon production services, driving multi-million dollars revenue across search/ads/personalization orgs.• Publish 20+ research papers on top conferences like NeurIPS, ICML, KDD, ACL etc. Presented invited talks in Amazon ML conferences.

Education

• Renmin University of China

  Computer Science

• Binghamton University

  PhD, computer science

  2013

Skills

• Machine Learning
• Truth Finding
• Data Quality Analytics
• Java
• Algorithms
• Distributed Systems
• Python
• Sql
• C
• Mysql
• Oracle
• Databases
• Eclipse
• Networking
• Natural Language Processing
• Knowledge Discovery
• Hadoop
• Software Development
• Mapreduce
• Programming
• Agile Methodologies