South KoreaRetail and e-commerce2025 – ongoingMusinsa
AI-native engineering organization and hiring
Musinsa reorganized its engineering work around AI tools and began hiring "AI-native" junior developers
Problem
Musinsa runs the Musinsa, 29CM and Soldout fashion platforms, with more than 17 million members and up to 3.3 million daily users during large sales. Its technology division wanted engineers who solve core work with AI tools and a development process that keeps pace as those tools change.
Approach
In January 2026 Musinsa opened its first large junior engineering intake since 2022, aimed at "AI-native" developers: no résumé was required, and successful candidates join a six-month internship before a full-time offer. Inside the technology division, teams use AI to check new builds against earlier builds and planning documents, to turn design files into interface code for four platforms, and to give several coding agents a shared knowledge base about the company's business.
Results
Takeaway
Musinsa changed hiring rules and engineering practices. In its internal test, the shared domain knowledge base improved answers to hard questions most, while average judged quality barely changed; Musinsa reported no company-wide productivity result.
Categories
- Industry
- Retail and e-commerce
- Business function
- IT and R&DHuman resources
- Innovation type
- Organization and ways of working
- AI technology
- Generative AIAI agents
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Company's own claim
Sources
Source summaries
Byline Network reports Musinsa's announcement of an open hiring round for "AI-native" junior software engineers under its Musinsa Rookie programme, the first large junior intake since 2022. Applicants need no résumé, sit coding tests and interviews, and join as six-month interns with a path to full-time roles. It cites more than 17 million members across Musinsa's platforms and quotes CTO Jeon Jun-hee on using AI as a tool.
A Musinsa engineer describes a two-layer domain knowledge base, an industry-standard core plus company-specific rules, that several coding agents read before answering questions about legacy code. In a test of 150 questions, keyword coverage rose from 67% to 81% and citations of standard concepts from 1.6 to 7.6 per answer, but the average AI-judge score moved only from 7.0 to 7.2. The author warns that answers about current behaviour still have to be checked against the code.
Musinsa announced its September 2026 tech conference, themed on a year of building products with AI. It lists examples from the technology division: AI that checks new builds against earlier builds and planning documents before QA, generation of interface code for four platforms from a design file, and a merged company-wide data query platform. It gives no productivity figures.
Full details are in the original sources.
