About KRAFTON AI Research Division
KRAFTON AI Research Division collaborates with various internal and external fields to provide AI solutions for diverse problems and develops our own services through proprietary deep learning technology research. Our direction broadly covers two areas:
- Increasing game production efficiency and universality through deep learning technology development.
- Developing Co-Playable Characters (CPC) that can play games alongside users.
About the AI Research Team
The AI Research team's research begins with a deep understanding of the operating principles of Large Language Models (LLMs). Based on this, we create core technologies that enable game production efficiency and new play experiences. We provide an environment where postdoctoral researchers can deeply focus on LLM research.
- We theoretically and empirically analyze the expressiveness, reasoning, and generalization mechanisms of LLMs, seeking new perspectives and algorithms to enhance efficiency, stability, and performance.
- We advance Language Model-based Agent technology to implement Gaming Agents that autonomously understand game environments and interact naturally with users.
- We expand our research scope to the Physical AI domain, envisioning the next stage of interactive intelligence.
- We extensively cover key LLM topics such as scaling laws, reasoning, tool calling, agentic RL, efficient training, multi-modal grounding, omni-model, and world models.
Culture Fit
Members of the AI Research Division interact and collaborate with team members from various fields through diverse projects, proposing creative ideas for various problems. An atmosphere where opinions are freely expressed, regardless of age or position, is encouraged. Teams are composed of individuals from diverse cultural backgrounds, and active support for overcoming language barriers through translation and interpretation is provided to facilitate lively communication. The AI Research Division aims to create an environment where researchers can focus on in-depth exploration. We possess the research infrastructure to conduct large-scale LLM training and experiments, and actively support external research activities such as conference participation and presentations. We provide an environment where researchers can share their expertise and grow together.
About the Role
KRAFTON is seeking a Postdoctoral Researcher to co-lead research into a deep understanding of LLM operating principles and performance enhancement. As a Postdoctoral Researcher, you will:
- Theoretically and empirically analyze the operating principles of Large Language Models (LLMs), and propose and verify new analytical perspectives and algorithms to improve model efficiency, stability, and performance.
- Explore new capabilities of LLMs such as tool use, agentic learning, and reasoning, and verify and develop them in interactive environments.
- Explore the potential of LLMs in the Physical AI and World Models domains, creating new forms of interactive intelligence.
- Collaborate with various research areas within the division to ensure research outcomes translate into actual game production and player experiences.
- Lead in-depth research within the team's research areas, leveraging your interests and strengths. The research conducted will be compiled into Tech Reports or published as papers/workshops at major AI conferences such as NeurIPS, ICLR, ICML.
Requirements
Required Qualifications
- Ph.D. in AI, Computer Science, Statistics, Electrical Engineering, or a related field, or expected to obtain before joining.
- Experience in theoretical/empirical analysis or modeling research of Large Language Models (LLMs).
- Deep understanding of factors influencing LLM performance, such as model architecture, training algorithms, and optimization techniques.
- Experience in research and experimentation using Python and deep learning frameworks (e.g., PyTorch, JAX).
- Quantitative analysis skills, including model performance evaluation, benchmark design, and experimental automation.
- Excellent communication skills for work.
Preferred Qualifications
- Possession of outstanding research achievements such as first-author papers published, presented, or awarded at top-tier AI conferences (e.g., NeurIPS, ICLR, ICML).
- In-depth research experience on LLM operating principles, such as internal structure analysis, scaling laws, representation learning, and model interpretability.
- Experience in training or analyzing Large-scale Language Models.
- Possession of a personal research vision and agenda for a specific LLM research area (e.g., scaling, reasoning, interpretability).