LLM Inference Engineer

The role

LLM Inference Engineer will optimize large language model inference performance and develop engines to stably serve EXAONE models on various frameworks. The role involves contributing to open-source LLM serving frameworks and designing NPU/GPU based serving strategies.

What you will do

• Derive and design improvement plans to maximize LLM inference performance.
• Develop engines to stably serve EXAONE models on various frameworks.
• Respond to feature improvements, bug fixes, and performance optimization requests via official channels of open-source LLM serving frameworks like vLLM, SGLang, and Ollama.
• Apply and improve chat_template to ensure LLM operates correctly in various languages and environments.
• Design NPU/GPU based serving strategies and research and implement model inference acceleration algorithms.

Required qualifications

• Experience in model inference acceleration and Inference Back-end implementation
• Experience in NPU/GPU based serving strategy and development
• Experience contributing to open-source LLM serving frameworks such as vLLM, SGLang, or Ollama

Nice to have

• Understanding of machine learning model lifecycle including development, training, serving, and deployment
• Proficiency in AI and machine learning technologies such as Python, C++, TensorFlow, PyTorch, ONNX, CUDA, and NPU
• Deep understanding of Transformer-based models and LLM structures and principles
• Understanding of LLM decoding strategies and inference technologies such as Greedy Search, Beam Search, and Top-k Sampling
• Experience using chat_template to design and apply for various environments

Before you apply

• LG AI Research is not a military service designated company, so recruitment of technical research personnel and transfers is not possible.
• The selection process may change.
• If you pass the document screening, the selection process will be guided separately.

About the company

Platform&Infra Team builds MLOps pipelines to optimize and manage the AI model lifecycle from development to service operation. They also perform security enhancement, infrastructure management, and resource optimization to support stable AI service operation.