Built by data engineers, for data engineers.
In 2022, Wei Tan left a data infrastructure role after watching the same category of failure repeat itself: schema drift at 3am, metric definitions that diverged across dbt and Looker, ETL jobs that required daily babysitting. He built the tool we kept wishing existed.
Why we built this
Loomkindle was founded in Seattle in 2022 by Wei Tan. Before starting the company, Wei spent four years building data infrastructure at a logistics analytics company in the Pacific Northwest — mostly Snowflake pipelines, dbt models, and a lot of on-call rotations when things broke.
The pattern that finally pushed him to build something: their revenue metric had three different answers depending on which tool you asked. The dbt model, the Looker dashboard, and the data science notebook all computed it differently — using column definitions that had silently diverged over 18 months of schema changes. Nobody noticed until a board meeting.
The existing semantic layer tools — dbt Metrics, Cube.dev, AtomicData — solved part of the problem. You could define a metric once. But they were all passive: define, then query. None of them watched your upstream schema and adapted. When a source column got renamed or a table got dropped, you still found out at 3am via a failing pipeline.
Loomkindle is the agentic version of that stack. We're not replacing dbt — we sit on top of it, and on top of your warehouse. We're not a BI tool. We're the layer that makes your existing definitions durable when the data underneath moves.
Small team, sharp focus
Four people building a product we'd want to use ourselves. We're hiring.
Wei Tan
CEO
Founder. Four years of Snowflake-based data infrastructure before starting Loomkindle in 2022. The schema drift and metric divergence problems Loomkindle solves are ones Wei lived first-hand.
Priya Nair
Senior Data Engineer
Distributed transform scheduling and Snowflake query pushdown optimization. Owns the connector ecosystem.
Marcus Webb
AI Infrastructure Lead
Builds the agentic routing layer and schema drift classification engine. ML ops and LLM workflow background.
Seo-Yeon Park
Product Designer
Developer experience — docs, CLI output, and dashboard UI.
How we build
Everything in code
No GUI-only workflows. Semantic models live in git. Infrastructure decisions are reviewable, diffable, and rollback-ready.
Practitioner-first
We build for the senior data engineer who has strong opinions about DAGs, argues about dbt vs SQLMesh on Slack, and reads schema migration notes before merging. No GUI-only workarounds, no magic black boxes.
Trust through transparency
Every routing decision is logged. Every schema change is tracked. Data teams should be able to explain exactly why their numbers changed.
Want to work with us?
We're a small team with two open roles. Or just reach out to talk data infrastructure.