Wei Niu

Lead agentic customer experience science team to research and develop best in class AI agent.

Lead and manage a team of machine learning scientists towards researching, developing and advancing key ML/AI initiatives for Amazon Connect, a disruptive cloud contact center service. Recent focus: Agentic AI for real-time agent assistance and self-service, leading agentic design, measurement and optimization. https://aws.amazon.com/connect/q/ https://aws.amazon.com/connect/wisdom/ https://aws.amazon.com/connect/contact-lens/ https://aws.amazon.com/connect/optimization/

Launched the demand forecasting solution for Amazon Connect Optimization and real-time agent assistant solution for Amazon Connect Wisdom.

ASR Built the state-of-the-art deep learning based ASR system, with application in phone audio transcription and realtime voice bot. Launched models to transcribe hundreds of millions/year conversational audios in multiple languages (English, French, German, Italian, Japanese, Hindi, etc). - voice activity detection - sequence-to-sequence ASR - multilingual ASR - language modeling NLU - neural dialog system - name entity recognition Customer Intent Prediction - Launched customer contact issue prediction model Customer Experience Personalization - Launched personalized help topic recommendation on Amazon Help Gateway

Recommender Systems/Geo-Social Data Mining/Machine Learning/Deep Learning

CSCE 470 Information Storage and Retrieval CSCE 438 Distributed Data-Intensive Systems

● Work at Transaction Risk Management System ML Team. ● Proposed and developed a mulitimodal sequential deep learning model for fraud detection. ● Trained and tested the model over 12 million clickstream sessions using GPUs. ● Achieve remarkable improvement over current production model which leads to an estimate of millions dollar potential saving from chargeback.

My research interest lies in large-scale data mining, applied machine learning/deep learning, and recommendation system. My dissertation focus on enhance existing and future geo-social systems through a systematic study of the intrinsic mutual reinforcement relationship between geography (geo) and user behaviors (social) through three important use cases: geo-spatial profiling, ranking and recommendation.