Our Products

Research-driven AI systems for operationally complex business environments — built from evaluation, routing, and agentic systems research.

Furiwake

AI Model Intelligence Platform

Live

Furiwake classifies, routes, and continuously evaluates AI model performance — so every task hits the right model, and you know it is working. Paired with Rensei, our benchmark-assurance layer used by AI labs and enterprise engineering teams to build defensible, reproducible evaluations.

Key Capabilities

  • ✦Smart model routing across LLMs — cost, latency, quality, and privacy aware
  • ✦Rensei benchmark assurance: multi-model ensemble judging with leave-one-out consensus (no single-judge bias)
  • ✦Rensei policy simulation: stress-test evaluation criteria via factorial grids and ANOVA before publishing
  • ✦Continuous drift detection with calibration gates — alerts when model rankings shift
  • ✦Reproducible experiment fingerprinting — re-run any evaluation months later and match results
  • ✦Annotation-to-benchmark pipeline with disagreement-tier routing for human review
Visit Furiwake

RL Insights Engine

Reinforcement Learning Business Intelligence

Coming Soon

Enterprise-grade RL platform that learns from your business decisions and provides adaptive, real-time optimization recommendations.

Key Capabilities

  • ✦Multi-armed bandit optimization
  • ✦Dynamic pricing and inventory optimization
  • ✦Customer journey optimization
  • ✦Real-time decision APIs
  • ✦Custom reward function design
Join Waitlist

Have a Custom Challenge?

We build bespoke AI systems for operationally complex domains. Let's discuss how our research-driven approach can help your business capture intent and bridge the execution gap.

Schedule a Consultation