real-company-interview-ai-coding-projects README

The project's README from github.com/CHENG-LIANG1/real-company-interview-ai-coding-projects. As one markdown file: README.md.

真实公司面试 AI Coding 项目

本仓库整理了三道来自真实招聘流程的 AI Coding 项目题,并给出一套可以复用于限时笔试、Take-home Assignment 和现场结对开发的通用方法。

为保护隐私,本文做了以下处理:

三道项目题

编号 项目 形态 主要考察 我的个人解法
01 Computer Use Agent Dashboard Web + Agent + 隔离桌面 Computer Use、工具循环、会话隔离、可观测性、运行时可靠性 ai-agent-dashboard
02 React Native 截图行动 Agent React Native / iOS,48 小时 多模态理解、结构化 Action、Human-in-the-loop、原生工具、Memory ContactFlow
03 Jira 风格任务管理系统 Web,一日 MVP 需求取舍、CRUD、看板拖拽、本地持久化、时间线、交付质量 ForceTrack

三个解法仓库是我针对对应题目的个人实现,包含源码、设计文档、测试与实际交付记录。本仓库继续只保留题目复盘和方法论。

ForceTrack:真实 AI Coding 对话历史

查看 ForceTrack 从 PRD 到 Repo Wiki 的 24 个真实对话轮次

通用方法

AI Coding 项目通用解法 将整个过程拆成:

  1. 需求澄清与范围冻结;
  2. Starter / 现有仓库审计;
  3. 数据、状态、工具与运行时设计;
  4. 垂直切片和 AI 任务拆分;
  5. 人工 Review 与自动化验证;
  6. 限时取舍、演示和最终交付。

三道题共同在考什么

三道题表面上分别是 Agent、移动端和项目管理系统,实际都在判断候选人能否:

真正拉开差距的通常不是代码量,而是:需求理解、架构边界、过程控制、验收证据和取舍能力。

阅读顺序

如果你准备类似笔试,建议先读三道题,再读通用解法。实际做题时,先复制通用解法中的需求矩阵、任务合同和交付检查清单,再根据题目删减。

说明

本文用于个人复盘、求职准备和工程学习。不同招聘流程的实际题面、时限与保密要求可能不同,请始终以你收到的正式材料为准。

real-company-interview-ai-coding-projects (CHENG-LIANG1/real-company-interview-ai-coding-projects)

project-management
Searching Database…

What is this?A write-up of three anonymized real-company AI coding interview take-home projects (an agent dashboard, a React Native action agent, and a Jira-style task tool), plus a general method for tackling timed take-home assignments and pairing interviews.
Is this the actual solution code, or just the problem write-ups?Just the write-ups and methodology; the actual solution code for each problem lives in its own separate linked repository (ai-agent-dashboard, ContactFlow, ForceTrack).
Are these the original companies' exact problem statements?No. The problems were anonymized and restructured, keeping only the technical goals, constraints, and acceptance criteria, not a verbatim copy of the originals.
What are the three example projects about?A Computer Use Agent Dashboard (web + agent + isolated desktop), a 48-hour React Native screenshot-action agent (multimodal, human-in-the-loop), and a one-day Jira-style task board MVP (CRUD, kanban drag-and-drop, local persistence).
What is ForceTrack's conversation history for?It shows 24 real dialogue turns from the initial PRD to the repo wiki for the Jira-style task-board project, illustrating how the author actually worked with AI during that take-home.
Can I see the actual code for the solutions?Not in this repo. Each solved project has its own separate repository (ai-agent-dashboard, ContactFlow, ForceTrack) linked from the comparison table.
Size
<0.1 MB
Stars
358
Updated
CommitAug 26
Age
1mo
Commits · 90 Days