[AI Frontier Div.] AI Inference Platform & Operations Lead (10년 이상)

Seoul; Seoul, South Korea. Posted 10 days ago

The role

The AI Inference Platform & Operations Lead will design, build, and operate serving platforms for proprietary and open-weight models. This person will lead the technical direction for the inference fleet, manage GPU resources, and establish operational standards for the AI Frontier division.

What you will do

Own the inference fleet and control plane
Design and operate serving platforms for proprietary and open-weight models
Select serving engines and orchestration layers
Establish technical roadmaps and operational standards including autoscaling and multi-tenant policies
Recruit and develop the team
Act as a single point of contact between AI Research, IT, and business squads
Build self-service tools and guides for model onboarding
Define SLOs and error budgets with business squads
Design GPU scheduling, allocation, and capacity planning models
Improve serving systems through continuous batching, prefix caching, and other techniques

Required qualifications

10+ years of experience in infrastructure, SRE, or platform engineering for large-scale traffic services
Experience leading teams responsible for critical-path production systems
Ability to select core technologies and explain trade-offs regarding performance, cost, and complexity
Experience defining and improving latency, availability, throughput, and cost metrics
Ability to negotiate SLOs and costs with business organizations and executives
Experience designing and operating large-scale distributed systems in Kubernetes or similar environments
Experience deploying and operating LLM, ML models, or high-computation real-time systems in GPU production environments
Experience operating capacity planning, quota, and resource isolation policies
Proficiency in writing production service code with Python or Go
Ability to read and modify large-scale open-source system code
No disqualification for overseas business trips

Nice to have

Experience operating LLM serving engines like vLLM, SGLang, TensorRT-LLM, or Triton Inference Server
Experience operating distributed inference orchestration or inference control planes like NVIDIA Dynamo, llm-d, KServe, or Ray Serve
Experience implementing serving features like continuous batching, prefix caching, or KV cache offloading in production
Experience applying optimization features like quantization checkpoints or speculative decoding in production
Experience operating multi-node GPU environments and GPU resource management tools
Experience with real-time voice/agent systems or real-time streaming serving for multimodal models
Experience optimizing low-latency pipelines using WebSocket or SSE
Experience serving agent workloads or operating LLM gateways
Experience building and recruiting new teams or platform organizations
Experience integrating researcher-made models into production

Before you apply

The hiring process may include a phone interview, assignment, personality assessment, technical fit interview, and culture fit interview
A 5-month probationary period applies
Hiring may close early upon completion of recruitment

About the company

Krafton's AI Frontier division connects AI technology to actual products and services, verifying and expanding new business opportunities.