Shivini · The studio behind Kinode
Shivini is an independent engineering studio and R&D lab. Kinode is a distinct AI infrastructure product developed within its engineering work.
Beyond one product
How does work run across machines that were never designed to cooperate? How can orchestration be trusted when steps fail? How do tools keep developers productive, and how do AI systems stay understandable to the people using them?
| Project | Distributed execution | Trustworthy orchestration | Developer tooling | AI systems | Human interaction |
|---|---|---|---|---|---|
| Kinode | Distributed execution | Trustworthy orchestration | Developer toolingNot a focus | AI systems | Human interactionNot a focus |
| Lycia | Distributed execution | Trustworthy orchestration | Developer tooling | AI systemsNot a focus | Human interactionNot a focus |
| HumanTick | Distributed executionNot a focus | Trustworthy orchestration | Developer tooling | AI systemsNot a focus | Human interactionNot a focus |
| Fabric | Distributed executionNot a focus | Trustworthy orchestration | Developer toolingNot a focus | AI systems | Human interaction |
| LoomAI | Distributed executionNot a focus | Trustworthy orchestrationNot a focus | Developer toolingNot a focus | AI systems | Human interaction |
| LoomCode | Distributed executionNot a focus | Trustworthy orchestration | Developer tooling | AI systems | Human interactionNot a focus |
Kinode benefits from that wider experience across distributed systems, AI, developer tooling and experimental products, while keeping its own scope, stage and evidence.
Selected work
Heterogeneous AI infrastructure
A local-first distributed model and inference service for heterogeneous, owner-authorized resources.
Boundary Support is tied to specific model, runtime, workload and device evidence. Broad hardware compatibility and full product acceptance are not claimed.
Distributed messaging framework
A message-driven saga framework for .NET, built around orchestration, durable state and explicit recovery.
Boundary Delivery is at-least-once, so handlers and external side effects must be idempotent. It is not an exactly-once guarantee.
Protocol and reference tooling
An issuer-bound creation-event commitment protocol for digital assets, with signed declarations and deterministic verification.
Boundary A signed declaration is not proof of human authorship or factual truth, and HumanTick does not detect AI-generated content.
Verified AI-composed interfaces
A Service-to-UI reference implementation that checks provider rules across adaptive Web, ChatGPT and Voice experiences.
Boundary A flight-domain slice with mock provider operations and recorded model output. Not a production service, and separate from Kinode’s compute fabric.
Local-first AI product
A personal AI web of reusable conversations, references, branches and inspectable context.
Boundary Bundled model runtimes, cloud sync, Windows or Linux packages and a notarized macOS release are not claimed.
AI coding orchestration
A local-first orchestration system that connects conversational planning with a deterministic coding execution engine.
Boundary Acceptance covers documented, bounded software gates. Hardening is pending and general production readiness is not claimed.
A portfolio of engineering work, not a product catalogue. Listing a project here does not make it a commercially available Shivini product. Stages and boundaries restate each project’s page on shivini.ai, reviewed November 2026.
Engineering philosophy
Failure belongs in the first architecture discussion. Recovery paths are designed, not discovered.
Prove the hard part early, then build the system around it and make it operable.
Claims stay within what the evidence supports, and stated limits travel with the result.
Contracts, data models and failure behavior are chosen deliberately, and people stay in control.
The studio behind KinodeDiscover Shivini (opens in a new tab)Services, projects and engineering notes at shivini.ai. Founded and led by Gökay Okutucu.