At MyShell, I test web and mobile AI Agents, including workflows, LLM behavior, multimodal components, and prompt safety. I helped resolve more than 30 logic defects and improved core workflow execution success from 85% to 99%.
I defined evaluation criteria for factual accuracy, relevance, consistency, format compliance, and latency. I validated more than 80 multimodal scenarios and established reusable evaluation baselines.
I built API and UI automation with Python, Pytest, and Playwright, and used Cursor and Playwright MCP to reduce API test script development time from about 30 minutes to 5 minutes. I also added prompt-injection and sensitive-information exposure scenarios to release regression testing.
Earlier, at Citi, I built a Cypress UI automation framework integrated with Jenkins and guided a three-person QA team. At JD.com, I tested e-commerce journeys and improved API automation reliability; at NetEase Games, I tested mobile game releases including Onmyoji.
