June 2, 2026
EnableFusion Inc. (hereinafter "EnF") and BioNexus Inc. (hereinafter "BioNexus") signed a joint research agreement on June 2 to develop "Fusion Co-Scientist", an AI system specialized for the fusion energy field.
Fusion Co-Scientist is a multi-agent AI system that enables fusion researchers to collaborate with AI to generate and validate research hypotheses. When a researcher presents a research goal, the AI explores relevant literature, multiple agents generate, critique, and evolve hypotheses, and the physical validity is verified through simulation—automating this entire sequence of processes.
This agreement is an equal joint research partnership combining the strengths of both companies. It aims to build an AI research system specialized for fusion by combining the AI Co-Scientist technology that BioNexus has validated in the biotechnology field with EnF's fusion domain expertise and GPU infrastructure.
The two companies began working-level discussions in February of this year and conducted preliminary joint development over approximately four months. As a result, by the time the agreement was signed, they had already collected about 60,000 fusion research papers to build a domain-specialized vector database, and completed a foundational RAG (Retrieval-Augmented Generation) system integrated with an LLM. In addition, they completed the development of six AI agents—including hypothesis generation, critique, ranking, and evolution—bringing the basic Co-Scientist pipeline to a working state.
In this joint research, BioNexus is responsible for AI system development (AI Core). It leads the design of the multi-agent architecture and the development of the hypothesis generation, ranking, and evolution agents.
EnF is responsible for fusion domain development (Domain Core). It leads the design of the collection and classification framework for fusion training data, the development and integration of the Fusion Simulator, the development of domain-specialized agents such as the Physics Agent and Simulation Agent, the training of the LLM and embedding models, and the establishment of an evaluation framework through a network of fusion experts.
The two companies will jointly develop the multi-agent system by dividing the work, aiming for a complete system composed of a total of nine agents.
This joint research will be carried out in three phases over a total of 24 months, from April 2026 to March 2028.
In Phase 1 (months 1–4), a prototype will be developed; in Phase 2 (months 5–12), the simulator integration and hypothesis validation framework will be established; and in Phase 3 (months 13–24), empirical research and a foundation for commercialization will be secured. The ultimate goal is to complete and commercialize the fusion-specialized Co-Scientist v1.0 system.
Fusion energy is drawing attention as humanity's ultimate energy source, yet many challenges remain to be solved, including plasma control and the optimization of nuclear reactions. Fusion Co-Scientist combines AI's large-scale literature analysis capabilities with fusion physics simulation, helping researchers approach these challenges more quickly and creatively.
Through this partnership, EnF plans to secure a leading position in AI-based fusion research platforms and to build technology assets that can contribute to the broader fusion research ecosystem in the future.