<- Back to projects

Case Study

Backend Interview Guide

Korean backend interview guide organized across database, cloud, system design, and programming, drafted and reviewed with Codex, Claude Code, and an agent harness.

Codex Claude Code Agent Harness Markdown Python GitHub

Project Overview

Backend Interview Guide is a Korean knowledge base for backend developer interviews. It is structured as a Markdown-first repository covering database, cloud, system design, and programming topics for real interview preparation rather than short Q&A snippets.

What Makes It Different

  • Agent-driven writing workflow: I used Codex, Claude Code, and an agent harness to draft, review, and refine documents in parallel.
  • Structured coverage: The repository currently spans 4 categories and 33 topic documents, with category indexes that keep the content navigable.
  • Interview-oriented writing: Each document focuses on how to explain trade-offs clearly in an interview, not just how to memorize definitions.

Content Scope

  • Database: scaling, optimization, transactions, caching, Redis, NoSQL, MongoDB
  • Cloud: container, Kubernetes, serverless, microservices, gRPC, service mesh, logging/monitoring
  • System Design: scalability, load balancing, high availability, API design, distributed systems, event-driven architecture, security
  • Programming: JavaScript, Go, data structures, algorithms, concurrency

Technical Challenges & Workflow

Challenge 1: Keeping a large Markdown repo coherent

The repository is split into category indexes and topic documents, so content quality depends on consistent structure, cross-linking, and topic boundaries.

Challenge 2: Making agent output production-ready

Parallel drafting is fast, but raw output is not enough. I used a review loop to normalize tone, remove weak explanations, and keep the final material interview-usable.

Challenge 3: Scaling content without losing maintainability

The repo includes runtime-discoverable category rules and a Markdown link checker so the documentation can keep growing without breaking navigation.

What I Learned

  • How to use multiple coding agents for content production without letting quality drift
  • How to turn backend topics into concise interview explanations with explicit trade-offs
  • How to design a documentation repo so agent workflows remain repeatable as the corpus grows

Impact

This project shows not only backend knowledge organization, but also how I build practical agent-assisted workflows with Codex, Claude Code, and harness-based parallel review.