Backend Software Engineer, Conversational AI - USDS

About the Team
The future of e-commerce customer service is intelligent, efficient, and AI-driven. Our team is dedicated to replacing traditional human-agent customer service with an advanced AI-powered conversational system that provides instant, intelligent, and seamless support for TikTok's global e-commerce platform. By leveraging Large Language Models (LLMs) and NLP, we are building an AI customer service system that can understand user queries, resolve disputes, guide transactions, and enhance the overall shopping experience without the need for human intervention.

Our cutting-edge AI is designed to handle complex customer interactions, including answering product inquiries, resolving order issues, processing refunds, and assisting sellers with operational tasks. Through LLM post-training, we ensure that our AI assistant is continuously learning and improving, providing more accurate, context-aware, and human-like interactions.

By joining us, you will be at the forefront of transforming customer service in e-commerce, helping build an AI system that understands, adapts, and provides intelligent solutions — all while reducing costs and improving efficiency for merchants and the platform.

In order to enhance collaboration and cross-functional partnerships, among other things, at this time, our organization follows a hybrid work schedule that requires employees to work in the office 3 days a week, or as directed by their manager/department. We regularly review our hybrid work model, and the specific requirements may change at any time.

What You Will Do
Develop AI-Powered Customer Service Systems: Design and implement an AI-driven conversational customer service agent that can handle e-commerce inquiries, complaints, refunds, dispute resolutions, and logistical issues, replacing traditional human customer service agents.
LLM Post-Training and Efficient Learning: Collaborate closely with algorithm teams to apply state-of-the-art LLM post-training techniques, such as instruction tuning, reinforcement learning from human feedback (RLHF), and continual learning, to optimize AI customer service responses with minimal labeled data.

Model Optimization and Efficient Deployment: Research and implement model compression, quantization, and inference optimization methods. Design efficient deployment strategies, enhance open-source frameworks, and explore cutting-edge generative AI to deliver world-class conversational agents.

RAG Technology and Applications: Lead the design and optimization of RAG frameworks, improve retrieval system performance, implement advanced algorithms, ensure system reliability, and build evaluation systems to optimize business outcomes.