Home › Platforms
Built once, licensed with source access.
Certain problems recur in every engagement, so we solved them once. Loonlake is a governed data foundation that sits on the warehouse a client already runs; Loonsong is the evaluation-first kit that takes AI from pilot to production. Both are accelerators that ship inside our engagements—licensed with source access, so that work begins from a working foundation rather than an empty repository.
The pair
Each stands alone. Together they form a way of working.
Loonlake makes the data trustworthy. Loonsong puts AI to work on it with the evaluation, guardrails, and review that make it safe to rely on—and each is licensed with source access and no dependence on our services.
Accelerator 01
Loonlake
Governed data foundation
The governed data foundation on which everything else stands.
When no one trusts the numbers, no one automates on them. Loonlake is an opinionated data foundation on the warehouse a client already runs: pipelines with tests and named owners, a modelled semantic layer with agreed definitions, lineage and documentation generated from the project, and access policies as code—assembled from proven open tooling rather than built from scratch.
Loonlake
What it includes
Pipeline and transformation templates with tests, alerting, and named owners
A modelled semantic layer: the objects, relationships, and definitions teams actually use
Lineage and a catalogue generated from the project, so they remain current
Access, retention, and residency policies as code, reviewable like any other change
Cost visibility by pipeline and by team, from the warehouse’s own usage data
LoonlakeHow it works
- SourcesERP · CRM · sensors · files
- Pipelinestests · owners · alerts
- Semantic layerobjects · relationships · definitions
- Lineage & accesspolicies as code
- Consumersanalytics · Loonsong
Runs in
On Databricks or Snowflake, within the client’s tenancy; other warehouses by agreement.
Accelerator 02
Loonsong
AI evaluation and production kit
Language models and agents, evaluated before they are trusted.
The model performs well in the demonstration; no one will sign off on it in production. Loonsong closes that gap: an evaluation harness built from the client’s own cases and run on every change, guardrail and refusal policies, a human-review queue for consequential decisions, and telemetry for cost, latency, and quality—on the client’s chosen models and gateway, within the client’s residency requirements.
Loonsong
What it includes
An evaluation harness built from the client’s own cases, run in CI on every change
A retrieval-grounded copilot reference application with citations, for the client’s team to extend
Guardrail and refusal policies, with a human-review queue and audit trail for consequential decisions
Cost, latency, and quality telemetry emitted to the client’s existing monitoring
Configuration for the client’s chosen model gateway and vendors—integrated rather than built
LoonsongHow it works
- Questionfrom a person or a workflow
- Retrievalgrounded in the semantic layer
- Model gatewaythe client’s gateway · any vendor
- Guardrails & evaluationyour own cases, every run
- Human reviewon consequential decisions
- Answerwith citations
Runs in
Within the client’s cloud tenancy, on the client’s chosen models; hosted open models where residency requires it.
Licensing
Licensed with source access.
Each accelerator is licensed with source access and the right to continue running it should the relationship end.
Support and upgrade agreements are sized to the client’s team, and upgrades ship on a published schedule.
The studio responsible for licensing is Platforms & Licensing.
Contact
From ambition to working software.
Whether you are accelerating AI adoption, modernizing a critical system, or building the capabilities to innovate faster, we bring the consultants who scope the work and the engineers who build it—the same people, from the first conversation. A senior engineer responds within one business day.
Get in touch