Chatsubo Labs

Chatsubo Labs was founded in 2000 by Kevin Raison — computer scientist, published AI researcher, and former Amazon engineer. For 25 years we’ve been building systems at the harder end of the spectrum: self-healing software, knowledge graphs, affective agent architectures, and autonomous physical systems.
We’re small by design. Our early work spans kernel-level systems programming, network engineering, and academic research — which means we can move seamlessly from a hardware schematic to a deployed ML model without a hand-off. Every engagement gets senior engineers and direct access to the person who will actually build the thing.

Member of:  AAAI  ·  ACM  ·  IEEE  ·  ISOC  ·  AAAC | Published at:  AGI-14  ·  JapTAL 2012  ·  SASO 2012  ·  GAMNLP-14

Three core disciplines. Every engagement is led by senior engineers — no hand-offs, no junior teams.

What We Build

Applied AI & Machine Learning

Custom model development, LLM integration, and end-to-end data pipelines — engineered for production, not demos. We’ve worked on problems in NLP, recommender systems, affective computing, and information retrieval since before ‘AI’ was a marketing term.

Robotics Systems

Embedded control systems, sensor fusion, and full-stack robot engineering from prototype to field deployment. We design for the real world — where hardware fails, environments are unstructured, and uptime is non-negotiable.

R&D Partnerships

Embedded research engagements for organizations that need experimental capability without building a full internal lab. We work best with clients who have a hard problem and the flexibility to explore unconventional solutions.

Selected Work

Open-source graph database & Prolog engine

Problem: Existing RDF stores lacked native Prolog reasoning and the performance needed for large-scale knowledge representation in production AI systems.
What we built: Designed and built VivaceGraph — a full graph database, RDF store, and Prolog implementation in Common Lisp. The codebase is open source and has been adopted by researchers in AI and knowledge engineering worldwide.
Outcome: Actively maintained; forked and used across multiple research institutions. Available at github.com/kraison/vivace-graph-v3.

Tags: Common Lisp · Graph Databases · Knowledge Representation · Prolog · Open Source

Affective agent architecture for goal-directed AI

Problem:  Classical AI planning systems lack the flexible, affect-modulated decision-making that allows agents to prioritize and adapt under uncertainty.
What we built:  Developed an affective agent architecture that uses emotion-inspired variables to modulate goal selection and planning behavior in multi-agent systems. Research was conducted in partnership with DePaul University.
Outcome:  Presented at the Seventh Conference on Artificial General Intelligence (AGI-14). Published in Springer Lecture Notes in Computer Science.

Tags:  AGI  ·  Affective Computing  ·  Multi-Agent Systems  ·  Planning  ·  Published Research

Automatic self-adaptation to mitigate software vulnerabilities

Problem: Critical infrastructure systems need to detect and respond to novel exploits in real time, without waiting for human intervention or a patch cycle.
What we built: Contributed FUZZBUSTER — a DARPA-adjacent research program developing self-organizing systems capable of detecting and automatically adapting to software vulnerabilities without operator intervention.
Outcome: Published at the IEEE Conference on Self-Adaptive and Self-Organizing Systems (SASO 2012). Contributed to the foundational literature on self-healing systems.

Tags: Cybersecurity · Self-Healing Systems · Self-Organizing Systems · IEEE · DARPA

Start a Project



Seattle, WA
206-801-5728
info@chatsubolabs.com