About this role
Join Trend ‧ Join New Generation
趨勢科技 - 全球雲端資安領航者 / 全亞洲最大軟體公司 / 企業版圖橫跨五大洲 / 趨勢全球研發基地在台灣
===============================================================
About the Role
We are building a production-grade, GPU-accelerated LLM serving platform that powers multiple AI products at enterprise scale. You will be responsible for designing, building, and operating the infrastructure that serves large language models — from raw Kubernetes cluster management to multi-GPU inference optimization and autoscaling.
Required Qualifications
Model Serving & Inference
- Operate multi-model LLM serving infrastructure
- Tune autoscaling policies to balance GPU cost and latency SLAs
Kubernetes & GPU Infrastructure
- Operate production K8s clusters with NVIDIA GPU nodes
- Handle GPU node lifecycle: NVIDIA driver setup
Infrastructure as Code
- Write and maintain Terraform/Terragrunt modules for AWS/GCP cloud
- Package platform components and model deployments as Helm charts
- Manage multi-environment configurations
Observability & Performance
- Maintain monitoring stack: Prometheus , Grafana ,
- Build dashboards for GPU utilization, KV cache occupancy, TTFT/ITL latency, and cost per token
- Set up alerting for SLA violations and OOM events
Bonus Skills
These are not required, but candidates with these skills will stand out.
- LoRA / PEFT fine-tuning workflows
- MLflow for experiment tracking, model registry, and automated adapter deployment
- Experience building LoRA adapter CI/CD pipelines (training → registry → serving)
- Experience with alternative inference frameworks such as SGLang or NVIDIA NIM , including deep Parameter Tuning for Continuous Batching , KV Cache management, and Speculative Decoding .
===============================================================
連結智慧 守護世界 --- Connected Intelligence for Securing a Connected World