[AI Frontier Div.] AI Inference Platform & Operations Engineer (5년 이상)
Salary not listed • Seoul; Seoul, South Korea • Full-time
Posted 10d ago
Remote Work Policy
On-site
Employment Type
Full-time
Categories
AI Infrastructure Engineer
About the job
We are seeking an AI Inference Platform & Operations Engineer to join our AI Frontier division. This role is crucial for connecting Krafton's AI technologies with actual products and services, and for validating and expanding new business opportunities. You will be responsible for building and operating the inference infrastructure that supports these products, ensuring fast and stable service connections for both proprietary and open-weight models. The mission involves maintaining consistent response times and availability under growing traffic, and efficiently managing GPU resources and inference costs. We foster a culture where opinions are valued regardless of rank or tenure, decisions are executed quickly, and failures are addressed through postmortems to strengthen systems. AI tools are integrated into daily development and operations, and we actively explore and apply new models and serving technologies.
Responsibilities
- Operate and improve the serving platform and control plane on Kubernetes-based GPU clusters, integrating serving engines like vLLM and SGLang.
- Implement request routing, auto-scaling, and multi-model operation functionalities.
- Automate model onboarding and deployment, including performance/quality regression testing, canary/rollback pipelines, and API gateway security.
- Establish and monitor Service Level Objectives (SLOs) for latency, throughput, availability, and error rates, analyzing bottlenecks and failure causes.
- Manage GPU resource allocation, isolation, and prioritization policies, and measure unit costs per model and workload.
- Implement and improve routing, scheduling, caching policies, and model integration code to meet service demands.
- Verify and apply optimization research results, such as quantization checkpoints and speculative decoding, in production environments.
Requirements
- Minimum 5 years of engineering experience, with at least 3 years in infrastructure, SRE, platform engineering, or ML systems.
- Experience operating production services where downtime directly impacts users.
- Designed and operated deployments, load balancing, scheduling, and auto-scaling in Kubernetes or similar orchestration environments.
- Built observability systems using metrics, logs, and traces, and structurally improved root causes of failures through on-call rotations and postmortems.
- Implemented operational automation tools or platform features in code and managed infrastructure/deployments reproducibly using IaC, GitOps, and CI/CD.
- Collaborated with cross-functional teams (model researchers, product developers) to align requirements and documented design/operational decisions.
- No disqualifications for overseas travel.