Research Engineer - Foundation Models

Responsibilities

• Design and implement end-to-end pipelines for training, evaluation, inference, and deployment of large-scale LLM and Multi-modal Foundation Models
• Implement training algorithms, data composition methods, optimization strategies, and evaluation methodologies to improve model performance and stability
• Lead the entire process of experimental design, implementation, execution, analysis, and documentation to verify research hypotheses
• Apply and improve model parallelism, data parallelism, pipeline parallelism, communication optimization, and memory optimization in large-scale distributed training environments
• Integrate and experimentally verify new model architectures and training/optimization techniques into existing codebases
• Build reproducible experimental environments to clearly analyze causes of model improvement
• Analyze and resolve issues such as performance degradation, instability, convergence failure, and distributed system errors during training
• Quantitatively analyze model performance, training efficiency, and system efficiency to support decision-making

Requirements

• Bachelor's degree or higher in AI, Computer Science, Statistics, Electrical/Electronic Engineering, or a related field
• Research or development experience in Machine Learning, Deep Learning, or Foundation Models
• Software engineering skills to understand and modify models, training loops, data loaders, and evaluation pipelines in large codebases
• Ability to quantitatively analyze experimental results and hypothesize causes for performance changes
• Strong communication skills for documenting and sharing complex implementation details and analysis
• Ability to structure and solve problems through collaboration with researchers, engineers, and product teams

Preferred Qualifications

• Experience with training, evaluation, and inference of large-scale LLM or Multi-modal Foundation Models
• Experience handling systemic issues in large-scale model development such as distributed training and model parallelism
• Experience analyzing and solving complex problems like training instability, performance bottlenecks, or data quality issues
• Experience designing or improving ML pipelines from training to deployment
• Understanding or experience in latest research topics such as LLM agents, reasoning, tool use, reinforcement learning, or multi-modal learning

About the Company

KRAFTON AI Research Division is building a next-generation game production paradigm centered on Large Language Models (LLM). We aim to increase game production efficiency and create new player experiences through deep learning technology and the development of Co-Playable Characters.