1. Home
  2. Jobs
  3. Materials Science Specialized LLM & AI Agent Development [A26-02]
Data & Computation

Materials Science Specialized LLM & AI Agent Development [A26-02]

National Institute for Materials Science (NIMS) · Tsukuba, Ibaraki (JP)

Advertised paySalaries will be determined based on ability, experience, performance, etc.
Apply on employer site →Posted September 2, 2026

About the role

Overview We will promote the R&D and implementation of a next-generation AI infrastructure to revolutionize materials development by maximizing the use of NIMS’s vast accumulation of materials science data, including scientific papers, experimental data/images, and crystal structures. Specifically, the candidate will be responsible for the continuous pre-training and fine-tuning of LLMs using our proprietary data, and the development of domain-specific tokenizers. Furthermore, leveraging frameworks such as LangChain, LlamaIndex, and AutoGPT, as well as the Model Context Protocol (MCP), you will design and build the architecture for "Autonomous Research AI Agents" that can independently operate and interface with NIMS's existing computational simulators and automated experimental systems. By designing advanced reasoning frameworks using cutting-edge AI technologies such as GraphRAG, along with hybrid AI systems that incorporate physical constraints, we aim to realize a data-driven research platform that dramatically accelerates the speed of materials discovery. For this position, we are seeking individuals who can apply AI technologies with intellectual curiosity toward the domain of materials science, and who find value in implementing the latest technologies from the IT industry into the research field with a proactive mindset. Qualification Deep understanding of Transformer architectures and hands-on experience in implementing and deploying deep learning models using Python and frameworks such as PyTorch or TensorFlow. Technical expertise in Large Language Models (LLMs), with practical experience in Continuous Pre-training, Fine-tuning, and RAG (Retrieval-Augmented Generation). Profound knowledge of software engineering and high-level engineering skills to design and build complex software systems from scratch. How to Apply Submit the specified application documents via e-mail. For more information and to apply, visit: A26-02 https://www.nims.go.jp/eng/employment/data/m4p8m5000000d060.html

Get new listings like this, weekly

We store your address to email you about Funded Science. Nothing else, and we don't share it.

Similar positions

← All science jobs