Research Engineer

JAPAN AI, a Geniee group company building JAPAN AI STUDIO as the core system of an enterprise, in which AI agents run workflows across all of a company's SaaS tools, is hiring a Research Engineer. The team aims to solve the problems that make today's agents give up: weak reasoning when searching across several SaaS tools, long-term memory over extended processes, multimodal handling, and low-latency inference for hundreds of companies at once.

The engineer conceives and compares agent harnesses (memory, context compression, communication between agents), designs quantitative benchmarks for large agentic tasks, supports automated evaluation of models and prompts, builds and tunes training-data mixes, and hands research results to the product engineering team. Publishing papers is not the goal; the work runs in production for about 200 companies.

Applicants need a master's or doctoral degree in computer science, machine learning or a related field, experience building complex LLM agent systems, large-model training and inference with PyTorch or JAX, a deep understanding of Transformers, and production-quality Python. Fluent Japanese or business-level English (either one) is required.