Unified Semantic Layer
Single source of truth for metrics and dimensions. Define once, consistent everywhere — in your warehouse, BI tool, and ML feature store.
Agentic ETL · Semantic Layer · Seattle
The semantic layer that routes itself.
Loomkindle is an agentic data semantic layer and ETL engine for analytics teams. Define your metrics once. Let the agent handle the wiring.
Analytics pipelines are still handcrafted.
Every data team we talked to has the same story: SQL transforms written by one engineer, owned by nobody, brittle to schema changes, impossible to trace. Metric definitions duplicated across dbt models, Looker LookML, and ad-hoc Python scripts — with no single source of truth. ETL jobs that require someone to check them every morning. When upstream changes a column name, you find out from a broken dashboard, not from a log entry. Loomkindle was built by the engineers who lived this problem for years before deciding to fix it.
> "The pipeline failed again. Third time this week because upstream renamed a column."
> "We have four different definitions of 'revenue' across dbt, Looker, and our data science notebook."
> "Every Monday morning I check if the overnight ETL finished. I've been doing this for two years."
How it works
Connect your sources
Point Loomkindle at Snowflake, BigQuery, Redshift, or DuckDB. Schema inspection happens automatically.
Define your semantic model
Declare metrics, dimensions, and entity relationships in YAML. No proprietary DSL — just structured config.
Agentic routing takes over
The Loomkindle agent monitors upstream changes, reroutes transforms, handles schema drift, and logs every decision.
Query anywhere
Your defined metrics work consistently in dbt, Airflow DAGs, Sigma, Hex, or direct REST API calls.
Capabilities
Built for the modern data stack
Agentic Transform Routing
The agent detects upstream schema changes and reroutes your transform DAG automatically — no manual intervention.
Pipeline Orchestration
Built-in scheduling and dependency resolution. Integrates with Airflow, Prefect, and Dagster — or runs standalone.
YAML-First Config
No GUI-only workflows. Everything is code. Your semantic model lives in git. Changes are reviewable, diffable, and rollback-ready.
Lineage & Observability
Column-level lineage out of the box. Every transform is traced. Every routing decision is logged. Monte Carlo and Great Expectations compatible.
Multi-Warehouse Native
Push-down query optimization for Snowflake, BigQuery, Redshift, and DuckDB. No one-size-fits-all translation layer.
Connects to your existing stack
Early feedback
What data engineers say
"We've been using Loomkindle in a private pilot. The schema drift handling alone saved us two hours of on-call pain last month — a Snowflake table rename that used to page everyone at 6am just got handled."
Data Infrastructure Lead — a 12-person data platform team at a logistics analytics company
"The YAML-first semantic model clicked immediately. We were already on dbt, so defining metrics in the same config pattern felt obvious. Took less than a day to migrate three critical definitions."
Analytics Engineering Manager — an analytics team at a growth-stage fintech
Stop hand-wiring your semantic layer.
Join data teams using Loomkindle to define metrics once and let the agent route them.