Skip to content
Akili
GitLab

Introduction

Mesh-centric Data Product Platform. Domain teams own their data products end-to-end. Declare in 6 YAML manifests, write business logic in SQL or Python, and the platform handles orchestration, execution, quality, serving, and governance.

Core Principles

Section titled “Core Principles”

Mesh-centric

Domain teams own their data products. The platform provides the self-serve infrastructure layer.

Declarative

Define your data products in YAML. Write your logic in SQL or Python — the platform handles the rest.

Multi-tenant

Complete tenant isolation at every layer — storage, messaging, compute.

Observable

Built-in quality scoring, lineage tracking, SLO monitoring.

Self-hosted

Runs on your own self-hosted infrastructure.

Encapsulated Logic

Section titled “Encapsulated Logic”

A data product is a self-contained unit of deployment. Internal logic — CTEs, staging views, helper functions, modular SQL files — are private implementation details. No external product can reference them. Consumers see declared outputs and quality contracts, nothing else.

Think of it like software engineering: the 6 manifest files (product, schema, quality, ingestion, serving, governance) are the public API. Everything inside logic/ is a private method. You can refactor internals freely without breaking downstream consumers.

Serving projections — star schemas, flat tables, caches — are disposable rendering concerns. They can be rebuilt, reshaped, or replaced without touching the core data product. The entity/event declarations, identity columns, and quality contracts are the only durable public interface.

Data Product Lifecycle

Section titled “Data Product Lifecycle”
flowchart LR
    DECLARE[Declare -- YAML] --> VALIDATE[Validate]
    VALIDATE --> REGISTER[Register]
    REGISTER --> DEPLOY[Deploy]
    DEPLOY --> EXECUTE[Execute]
    EXECUTE --> SERVE[Serve]
    SERVE --> MONITOR[Monitor]
    MONITOR -.->|Update + Redeploy| DECLARE

Architecture Overview

Section titled “Architecture Overview”

Akili consists of a control plane (API and services), an execution engine (orchestration), and a multi-tier serving stack.

flowchart TD
    PORTAL[Portal] --> CP[Control Plane]
    CP --> ENGINE[Execution Engine]

    CP --> PG[Structured Store]
    CP --> RP[Event Bus]
    CP --> CEPH[Object Store]

    ENGINE --> PG
    ENGINE --> RP
    ENGINE --> CEPH
Section titled “Quick Links”