Kakao CTO Jung Kyu-gyu-hosted panel talk: "Developers must evolve into orchestrators in the AI era"

At the if(kakao)25 Krew Day AI panel talk moderated by Kakao CTO Jung Kyu-gyu, external experts discussed changes in development paradigms and infrastructure strategies in the AI era. The panel included brad, Head of Google Search Engineering; Shin Jung-gyu, CEO of lablup; and robin from KakaoTech's AI Native Strategy Team.

Importance of AI Infrastructure and Hardware Control

Shin Jung-gyu introduced Backend.AI as a solution with a vertically integrated stack from device drivers to MLOps for container-based GPU resource management. He emphasized that precise hardware-level control is essential because the non-deterministic nature of GPU specifications and the PTX layer in the LLM era impacts performance.

Google's AI Transformation and Search Evolution

brad explained how Google reorganized its organization with an 'AI First' strategy after the crisis, fostering an engineering environment. He clarified that Google's 'AI Overviews' summarize and restructure search results using LLMs, advising that exposing raw data to LLMs without restrictions for refinement is crucial for maintaining timeliness when building internal AI search.

Changes in Developer Roles and Workflows

Shin Jung-gyu predicted that 40% of coding is currently replaced by AI, with this figure expected to reach 70-80% by the end of this year. He argued that developers must transform from writing code directly into 'orchestrators' who command AI agents in parallel.

robin shared how KakaoTech's AI Native team improved development speed by adopting a flexible workflow based on POC (Proof of Concept) instead of the waterfall model. He analyzed that as the use of AI tools becomes widespread, the ability of senior engineers to make quick decisions and coordinate will become the core of productivity.