Noeon Research is an ambitious deep-tech startup working on a novel natively agentic graph-neuro-symbolic system with general capabilities. We are an international 30-people team with headquarters in Tokyo, Japan.
We are developing Noeon – a novel natively agentic AI system with general capabilities based on carefully selected principles from mathematical category theory, knowledge representation, and computational linguistics. Our technology prioritizes interpretability, revisability, and generality, allowing for safe and efficient adaptation to a changing world.
At the core of Noeon’s approach is a typed, compositional graph structure expressive enough to capture expert intuitions, formal enough to compile directly into solver input, and structured enough for LLMs to learn and manipulate. Rather than relying on manual, brittle translations from problem descriptions to formal constraints, Noeon’s technology allows problems described in natural language to be automatically elaborated into formal ontologies, translated to solvers, and solved - with every step inspectable, auditable, and reversible.
AI Safety
We prioritize AI safety by focusing on interpretability in order to unlock capabilities safely. We want to be responsible in our research and development. We don’t turn our noses up at AI safety research and take the issue of alignment seriously.
What We Offer
- Competitive salary
- Flexible schedule
- High autonomy
- Low bureaucracy
- Medical allowance
- Language courses
Our Values
Intellectual Bravery — We confront established mainstream research, embrace uncertainty, and fearlessly innovate.
Explicit over Implicit — When changes or alternative approaches are needed, we communicate them upfront.
Focus — We are developing complex technology with limited resources. Essentials first; extras later.
Cohesion — Our team collaborates effectively, takes responsibility for shared goals, and can always rely on one another.