China’s AI Agent interview circuit, documented and open-sourced

What it does This repository is a static-site knowledge base that aggregates interview questions, real experience reports from platforms like NiuKe, and study roadmaps for AI Agent engineering roles at companies from Alibaba and ByteDance to OpenAI and Anthropic. It covers the full stack from Transformer theory and RAG pipelines to MCP, Agentic Coding, and production deployment concerns, with most questions accompanied by reference answers.

The interesting bit Rather than dumping links, the maintainer built an “evidence model” that scores question confidence by source count, timeliness, and whether it came from first-hand accounts. There is even an Interview Collector Agent that installs as skills into Claude Code, Cursor, and GitHub Copilot CLI to scrape public forums and structure new entries automatically.

Key highlights

• 300+ questions with reference answers across 15 company categories, plus 69 foundational “eight-legged essay” concept questions
• 16-week study roadmap and six timed hands-on challenges, such as building a ReAct Agent or an AI Code Review Agent
• Coverage of emerging 2026 topics like MCP, Agent Harness evaluation, and runtime resilience
• Public interview experiences indexed with an evidence-based confidence score
• GitHub Pages deployment with both classic and exam-prep layout variants

Caveats

• The repository is overwhelmingly Mandarin; English speakers will need translation tools
• Some company sections (Tencent, Baidu, Google DeepMind, Microsoft, and startups) list interview questions and requirements but lack real experience reports, so depth varies by employer

Verdict Engineers in China targeting Agent roles at major tech firms should bookmark this. It is a curated knowledge base, not a framework or library, so expect to read and memorize rather than import and run.