Baidu · ERNIE
LLM Experience Research Assistant
Supported experience research for ERNIE Bot’s NLP track, serving model and data teams. Through desk research, in-depth interviews, and competitive teardowns, I reconstructed real user–LLM interactions and turned fuzzy notions of usability into standards a model can learn and teams can evaluate. The work covered student LLM use, behavioral comparisons of models such as DeepSeek, and applications including general marketing, resulting in experience reports and product recommendations.
My role
Conducted competitive experience research and in-depth interviews with target users, produced insight reports and product recommendations, and helped translate user needs into LLM data standards.
Scenario-based research on student LLM use
- Combined desk research and in-depth interviews to study how university students use LLMs for reading, creation, and search, producing a scenario × behavior × need framework.
- Delivered experience reports and product recommendations, helping turn user experience into LLM data standards.
DeepSeek & competitor research
- Analyzed the value and problems of chain-of-thought behavior from real social-media use cases and user feedback, comparing DeepSeek, Kimi, GPT, and other models.
- Mapped chain-of-thought behavior across knowledge queries, analysis, problem solving, decisions, and creation, while connecting user-visible behavior to the underlying technical principles.
What I learned
For foundation models, user research turns a fuzzy sense of usability into standards that models can learn and teams can evaluate. That remains the foundation of how I approach AI products.
Next experience
Tencent · Hunyuan