Artificial Intelligence & Biomedicine Lab (AIBL)

Core AI Foundations · Secure and Efficient AI Systems · Biomedical and Digital Medicine AI

From Mathematical Principles to Deployable Intelligence

Lab Description: 

• The Artificial Intelligence and Biomedicine Lab (AIBL: https://aibl.snu.ac.kr/) at Seoul National University develops globally competitive Core AI foundations and systems and advances them toward high-impact biomedical and digital medicine applications.

• AIBL is distinguished by an integrated three-layer research framework that connects mathematical foundations, advanced AI technologies, and real-world translation.

1) Mathematical and Algorithmic Foundations of AI

AIBL investigates the fundamental principles underlying learning, generation, reasoning, and decision-making. Core research areas include:

• Probabilistic generative modeling, including diffusion models, flow-based models, stochastic processes, and likelihood-based and likelihood-free learning;

• Reinforcement learning, sequential decision-making, control, planning, and multi-agent learning;

• Statistical learning theory, probabilistic inference, optimization, representation learning, uncertainty quantification, and generalization;

• Theoretical foundations of multimodal learning, foundation models, agentic intelligence, and efficient inference; and

• Formal characterization of model assumptions, identifiability, convergence, robustness, computational complexity, and failure modes.

The objective is not merely to apply established architectures, but to formulate fundamental research questions and develop original theories, algorithms, and learning principles that can advance Core AI itself.

2) Core AI Models and Secure Systems

AIBL translates theoretical advances into broadly applicable AI models, platforms, and systems with independent scientific and practical value beyond any single application domain. Major directions include:

• Advanced generative AI and multimodal foundation models;

• Large language and multimodal models with reasoning, planning, tool use, and agentic capabilities;

• Cooperative and competitive multi-agent intelligence, coordination, communication, and credit assignment;

• Efficient and lightweight AI through model compression, quantization, distillation, parameter-efficient adaptation, sparse computation, and inference acceleration;

• Secure and privacy-conscious on-premise AI for restricted or sensitive data environments;

• Edge and on-device intelligence, distributed inference, and AI software–hardware co-design; and

• Reliable deployment under practical constraints involving privacy, security, latency, computational resources, and system robustness.

Through this layer, AIBL aims to establish globally competitive Core AI technologies that can stand independently as foundational methods and deployable systems, rather than serving only as components of biomedical applications.

3) Biomedical and Digital Medicine AI

AIBL further specializes and validates its Core AI technologies for challenging problems in biomedicine and digital medicine. Research areas include:

• Computational and medical imaging;

• Multimodal integration of imaging, text, clinical, pathological, genomic, and other biomedical data;

• Bioinformatics, computational biology, spatial and molecular data analysis;

• Protein structure, molecular representation and generation, and AI-driven drug discovery;

• Secure medical foundation models and on-premise biomedical AI systems; and

• New digital medicine technologies that may progress, where appropriate, toward real-world deployment and clinical validation.

This research framework establishes a continuous pathway from mathematical principles and original AI algorithms to globally applicable Core AI systems, and subsequently to advanced biomedical technologies. Clinical validation is treated as an important potential endpoint of technological translation rather than the sole focus of the laboratory.

Research and Training Mission

• AIBL aims to train researchers with distinctive expertise across the complete research continuum of theory, algorithms, implementation, systems, and translation. Members are expected to progress beyond the use of existing AI models by learning to:

• Formulate original and technically significant research problems;

• Understand and analyze the mathematical principles underlying modern AI;

• Reproduce, debug, modify, and extend state-of-the-art methods;

• Design theoretically grounded algorithms and reliable AI systems;

• Conduct rigorous theoretical and empirical validation; and

• Translate research outcomes into globally relevant Core AI and biomedical technologies.

• Researchers are supported in developing work suitable for publication at premier international AI and machine learning conferences and in leading journals across AI, biomedical AI, bioinformatics, computational biology, and digital medicine.

Research Environment

• The principal investigator, Professor Kyungsu Kim, is affiliated with the School of Transdisciplinary Innovations, the College of Medicine (Department of Biomedical Sciences), and the AI Graduate School (Department of Engineering, IPAI (Interdisciplinary Program in Artificial Intelligence)) at Seoul National University.

• Through an extensive collaborative network spanning AI, engineering, medicine, and biomedical science, AIBL provides opportunities to conduct interdisciplinary research using real-world medical and biomedical data in addition to publicly available datasets.

• Medical and biomedical AI involves challenges beyond those commonly encountered in general-purpose AI research, including restricted data access, privacy and security requirements, heterogeneous multimodal data, rigorous validation, and the responsibility associated with technologies that may affect patient care. AIBL provides the collaborative environment, technical infrastructure, and interdisciplinary expertise required to address these challenges systematically, including through secure on-premise and edge AI systems.

Role Description

• AIBL is recruiting full-time Master’s, Integrated Master’s–Ph.D., and Ph.D. students to work as AI researchers at Seoul National University in Seoul, South Korea.

• Researchers will conduct theoretical and empirical AI research, implement and extend state-of-the-art methods, develop original algorithms and systems, analyze complex data, and collaborate with experts across AI, engineering, medicine, and biomedical science.

• The position requires demonstrated competence in both of the following dimensions:

  • AI Implementation and Engineering Capability

• Applicants must demonstrate a strong foundation in computer science and AI, together with the ability to independently:

  • Understand and reproduce state-of-the-art AI methods;

  • Implement research ideas using modern AI frameworks;

  • Debug, modify, and extend existing models and codebases;

  • Design rigorous experiments and analyze results;

  • Adapt advanced methods to new research questions and operating environments; and

  • Build reliable and computationally efficient research prototypes.

  • Theoretical and Analytical Capability

• Applicants must also demonstrate the mathematical maturity and logical reasoning required to:

  • Understand the theoretical principles underlying AI models and algorithms;

  • Formalize research questions, assumptions, and objectives;

  • Follow and construct mathematical derivations, proofs, and rigorous arguments;

  • Interpret empirical findings beyond performance metrics alone;

  • Analyze identifiability, convergence, generalization, robustness, computational complexity, and uncertainty where relevant; and

  • Identify methodological limitations, hidden assumptions, and potential failure modes.

• Both dimensions are essential. Implementation experience without sufficient theoretical and analytical depth, or theoretical knowledge without the ability to implement, modify, and validate working AI methods, is insufficient for the research conducted at AIBL.

Key Responsibilities

• Conduct original research across AI foundations, Core AI models and systems, and Biomedical and Digital Medicine AI.

• Develop theoretically grounded methods in generative modeling, diffusion, reinforcement learning, multimodal learning, multi-agent intelligence, and related areas.

• Reproduce, critically analyze, modify, and extend state-of-the-art AI methods rather than treating existing models as black-box tools.

• Develop efficient, lightweight, secure, and deployable AI systems, including on-premise, edge/on-device, and software–hardware-integrated solutions.

• Design rigorous theoretical analyses and controlled empirical evaluations.

• Translate Core AI advances into high-impact biomedical and digital medicine technologies where appropriate.

• Present and publish original research at premier international AI conferences and in leading journals.

• Collaborate effectively with AI researchers, engineers, biomedical scientists, and medical experts.

Qualifications

• Interdisciplinary training in computer science or AI together with mathematics, statistics, physics, electrical and computer engineering, biomedical engineering, biochemistry, computational biology, chemistry, bioinformatics, medical imaging, pathology, or a related field is highly valued. A double major or equivalent interdisciplinary experience is preferred but not required.

• A strong foundation or demonstrated potential in computer science, artificial intelligence, machine learning, or a closely related field;

• Substantial interest or competence in mathematics, statistics, physics, theoretical computer science, optimization, or another discipline requiring rigorous quantitative reasoning;

• Proficiency in AI/CS-related SW language, together with experience using modern AI and deep-learning frameworks;

• The ability to independently read, understand, implement, and critically evaluate recent AI research;

• Strong motivation to pursue technically rigorous and original research spanning theory and implementation; and

• Relevant academic, independent, open-source, competition, or industry experience demonstrating the required capabilities.

Required Application Materials

• Applicants must submit an application by email with sufficient evidence of both AI implementation capability and theoretical or analytical capability. Claims of proficiency without concrete supporting evidence may not be sufficient for evaluation.

• The application must include:

  • A curriculum vitae;

  • An academic transcript;

  • A concise personal and research statement explaining the applicant’s background, research interests, and reasons for applying to AIBL;

  • Evidence of AI implementation and engineering capability, such as research code, GitHub repositories, reproducible project results, technical demonstrations, publications, patents, competition results, or technically substantive project documentation;

  • Evidence of mathematical and analytical capability, such as relevant coursework and grades, theoretical research, mathematical derivations, proofs, technical reports, publications, or other work demonstrating rigorous and independent reasoning; and

  • A clear description of the applicant’s individual contributions to each submitted project or research outcome.

• Evidence may come from academic research or from technically substantive industry experience. Applicants from industry are evaluated based on the originality, technical depth, implementation quality, and analytical rigor of their contributions rather than on academic publications alone.

• Applicants must not submit proprietary or confidential information. Industry experience may instead be demonstrated through non-confidential technical descriptions, sanitized code or results, patents, public demonstrations, and clear explanations of the applicant’s individual role and contributions.

• AIBL welcomes highly motivated individuals who aspire to advance the mathematical and algorithmic foundations of AI, build globally competitive Core AI technologies and secure systems, and extend these advances toward high-impact Biomedical and Digital Medicine AI.