FinOps built for Databricks
Lakemine finds where your Databricks spend is being wasted, ranks every finding by dollar impact, and tells you how to fix it. It installs into your own Databricks account, reads the usage and billing data Databricks already keeps, and writes its findings to tables you own.
Your usage data stays in your account. Lakemine holds no standing access to your environment at any point.
Start here
Install it
The two surfaces
Lakemine has two places you will look at, and they answer different questions. Keeping them straight makes everything else easier to follow.
| Where it lives | What it tells you | |
|---|---|---|
| Cost dashboard | A Lakeview dashboard inside your own Databricks workspace, named “Databricks Cost Optimization” | Where your money is going and what to fix. Cost by workspace, by day, by user, by tag, and the ranked list of findings. |
| Lakemine portal | app.lakemine.ai | Whether Lakemine itself is healthy and licensed. Your installs, their last check-in, your plan, and the aggregated totals each scan reported. |
Findings never leave your Databricks account. The portal sees aggregated counts and totals for licensing and billing, and nothing else. The exact field list is documented in What is Lakemine.
The analysis is rule based. Every finding comes from an explicit, documented rule against your usage data, with the savings arithmetic shown. There is no model in the loop, and no output you cannot trace back to a query.
What Lakemine looks for
Seventeen detection engines run on every scan, each one owning a single failure mode. They fall into five groups:
- Right workload, right compute. Jobs on all-purpose clusters, SQL workloads on interactive clusters, bursty classic warehouses that should be serverless.
- Sizing and utilization. Oversized instance types, autoscaling left off, idle clusters with weak auto-termination.
- Purchasing and runtime efficiency. On-demand instead of spot, outdated Databricks Runtime, Photon left off.
- Job and query efficiency. Repeated short job-cluster starts, always-on streaming, failing jobs, many small jobs that should be consolidated, queries spilling to disk.
- Governance and FinOps hygiene. Clusters with no policy, spend with no budget guardrail, high-cost resources with no chargeback tags.
Each engine and its remediation steps are documented in the Engines section.