大模型平台研发（训练推理稳定性工程） - TikTok内容安全平台
Core
Design and build stability systems for large model training and inference, ensuring high availability and rapid recovery for TikTok's international content safety platform.
Role type
Senior SRE / Stability Engineer (AI Infrastructure)
Builds
Large model training and inference platforms, stability tooling, and observability systems for content safety
Domain
AI Infrastructure / Cloud Native / Content Safety
Deliverable
infrastructure
Required skills
Python, Go, Java, C++, Linux, Kubernetes, container orchestration, distributed systems, observability (Metrics/Logging/Tracing), performance analysis, root cause analysis, chaos engineering, capacity planning
Preferred skills
Large model training/inference stability, GPU heterogeneous compute operations, vLLM/SGLang inference engines, AI/Agent for operations automation
Responsibilities
Design SLI/SLO metrics and lifecycle management for training/inference stability; Review architecture for high availability and disaster recovery; Implement change control, canary releases, and rollback mechanisms; Execute stress testing and chaos drills for major events; Build automated fault analysis and runbook execution tools