Enterprise BI Platform Migration
Replaced a six-figure proprietary BI stack with self-hosted Apache Superset — plus an AI query layer.
- Role
- Lead engineer & architect
- Client
- US fintech (anonymized)
- Year
- 2026
- Lane
- AI & Data Platform
The problem
A fintech's analytics ran on a proprietary BI platform with an annual license that scaled with every new seat and board. Leadership wanted the same dashboards, on the same Snowflake data, without the bill — and ideally a way to just ask questions in plain English.
What I built
A full migration onto Apache Superset: I rebuilt the CEO liveboards tab-for-tab and validated the numbers cell-for-cell against source tables. I reverse-engineered the old platform's private API to capture every tile, shipped a custom Superset image with Snowflake, SSO, and an MCP natural-language layer, and deployed the whole thing to the company's EKS platform with Terragrunt, Karpenter autoscaling, and Helm.
Outcome
Same analytics, zero license cost, running on infrastructure the company already owned — with an AI layer that lets non-analysts query the warehouse in natural language.