I'm an AI researcher and developer specializing in protein structure prediction and large language models. Currently serving as CTO at mefriend.ai and as a Visiting Scientist at KIAS, I combine expertise in PyTorch multi-GPU computing, kernel optimization, and cloud infrastructure. My work spans from deep learning research to production deployment, with accomplishments in protein language models, and AI system optimization.
Currently researching protein 3D structure prediction using Protein Language Model and diffusion models, developing rapid MSA generation methods, and implementing ensemble prediction approaches while managing cloud-based protein databases.
Spearheading cutting-edge research in language model algorithms and overseeing the deployment of machine learning operations. Focused on enhancing model performance and operational efficiency, contributing to the development of scalable AI solutions.
Leading the technical direction of an AI entertainment platform as CTO and co-founder. Fully responsible for the core inference system development and optimization. Oversee all technical aspects while also participating in overall strategic planning and vision as a founding member.
Developed EnsBolz, an ensemble generation tool utilizing Boltz. (selected as a speaker for the LA Biophysics Society 2025)
Developed plmMSA, a fast and diverse multiple sequence alignment tool utilizing a Protein Language Model. (Presented at CASP16 and LA Biophysics Society 2025)
Achieved 3rd place in the Antibody/Peptide category. Developed plmMSA, a fast and diverse multiple sequence alignment tool utilizing a Protein Language Model.
Led the development of AI speaker's inference system and complex intelligence components, focusing on optimizing model performance and implementing advanced conversational capabilities.
Served as Team Leader in the 2023 Samsung Computer Engineering Challenge, addressing the challenge of accelerating LLaMa inference. My role encompassed the leadership and direction of the team's efforts in developing and integrating Kernel Fusing techniques. I was also responsible for orchestrating the Scheduler to ensure efficient task allocation and execution. The project culminated with me leading the final evaluation to assess our project's performance and deliverables.
As the Lead Developer for a cafe management platform designed for cafe operators, I was responsible for the comprehensive development of the platform. This included constructing the database, building APIs, designing and implementing the frontend, and designing the network architecture for application serving.
As the Lead Developer for the SW Expert Training Program's Healthcare Project, I was tasked with implementing MLOps using TFX. My responsibilities included frontend design, database architecture and construction, API development, and deploying services using a Kubernetes environment. Additionally, I managed the network infrastructure setup and oversaw the overall development process.
Extensive experience in optimizing Large Language Models using PyTorch's distributed training capabilities. Specialized in kernel fusion techniques for maximizing GPU utilization and reducing memory overhead in complex deep learning architectures.
Expert in implementing end-to-end CI/CD pipelines using Docker, Kubernetes, and ArgoCD. Proficient in managing development workflows with Jenkins and Jira, while ensuring secure credential management through Infisical.
Deep understanding of AWS ecosystem, including EC2, S3, RDS, and Lambda. Experience in designing scalable and cost-effective cloud infrastructures for production environments.
Strong foundation in building responsive web applications using Next.js framework. Skilled in creating maintainable UI components with TypeScript and implementing modern designs through Tailwind CSS.
Proficient in developing RESTful APIs using FastAPI and Express.js frameworks. Experienced in designing and managing both SQL (PostgreSQL, MySQL) and NoSQL (MongoDB) databases, with expertise in Prisma ORM for type-safe database operations.
Selected for the 2025 DIPS Project (Leading AI Startup in Korea) (mefriend.ai) (Thanks to my team-mates!)
Selected as a speaker for the LA Biophysics Society 2025 to present research on complex ensemble generation using deep learning methods (Thanks to my advisor and co-workers!)
Won the SKKU Creative ICT Competition for developing an AI speaker with the naly.ai team
Selected for the Youth Entrepreneurship Training Program, a prestigious government-backed initiative supporting young entrepreneurs
Secured the 7th place in the Samsung Computer Engineering Challenge for accelerating inference in LLM models within a multi-GPU environment.
Achieved 1st place in the SW Expert Training Program for outstanding performance in a company-led project.
Won the Popularity Award in the SW Expert Training Program by securing a high number of votes in a company-led project.