Global AI Frontier Lab Seminar Series: Automating AI R&D: From Research Idea to Model Deployment

Speaker: Dr. Sung Ju Hwang

Location: TBA

Date: Monday, May 12, 2025

Location: 1 Metrotech Center (Floor 22)

What if AI could draft a research proposal, generate the code, run experiments, and deploy the model in an end-to-end manner? Our research group at KAIST and DeepAuto.ai is working toward this goal, building a pipeline of multi-agent systems that collectively serve as an AI researcher, engineer, and data scientist. First, our ResearchAgent initiates the process by generating original, implementation-ready research proposals through iterative literature analysis and review emulation. Then, Paper2Code translates these proposals into modular, runnable code by coordinating planning, analysis, and coding agents. Finally, AutoML-Agent takes over to handle model training, tuning, and deployment at scale, achieving strong performance without manual intervention. Together, these systems form a seamless pipeline for AI R&D: Research Idea → Proposal → Code → Model → API. We expect that the successful implementation of such an automated AI R&D pipeline will substantially accelerate the progress of artificial intelligence, enabling AI systems to autonomously improve and evolve, thereby bringing us closer to the emergence of artificial general intelligence sooner than previously anticipated.