Hakuya-owned systems
Production AI systems we run ourselves
We judge the work by what runs and what can prove it ran. These are systems Hakuya designed and operates for its own work. Named client engagements appear only after they are real and publishable.
01
Asgard operations fleet
Hakuya-owned specialist agents coordinate triage, code, review, research, and operations. Dispatch records and run history make each handoff inspectable.
02
Muninn decision memory
Hakuya's retrieval layer brings prior decisions and outcomes back into new work, with relevance and recency ranking instead of a growing prompt copied by hand.
03
Heimdall automation
Hakuya's scheduler runs recurring and event-driven jobs with health checks, bounded retries, and a clear route to a human when a job cannot recover.
04
Release guardrails
Tests, review gates, cost controls, and deployment evidence sit between a model proposal and a production change.