Backend & Data Platform: BigQuery, Cloud Functions, API

An online education company selling high-ticket courses through paid ads, a video sales funnel, booking calls and a closing team. Every number lived in spreadsheets with different definitions, so nobody could say how many leads showed up to calls last week or which ad paid for a sale. I was hired to build the system of record and the backend around it. Client anonymized.

Ingestion. Ten cloud functions in Python and Node on a scheduler pull from the booking calendar, the CRM, the funnel builder, the ads platforms and the payment providers, plus real-time webhooks with signature verification for bookings and payments. Everything lands raw in a BigQuery warehouse.

Modeling. Modeled views for leads, ads and payments rebuilt every 15 minutes with one definition per metric: shows, wins, unique leads, re-bookings, cost per lead, attribution by creative. An identity layer written in SQL matches the same person across emails and phone numbers in five systems.

API and dashboard. A FastAPI service with a React front end on Cloud Run behind Google sign-in: 12 views for operations, acquisition, team and finance, cached queries and per-view access control.

Reliability. 32 automated daily checks compare sources, catch silent failures and surface a status badge in the dashboard. Secrets in Secret Manager, idempotent writes, retries with dead-letter handling.

One source of truth for the founder, the media buyer and the sales team

Show-rate measurement corrected after finding the CRM did not update timestamps on status changes

Sales recovered in attribution by matching payments to people instead of emails

Dashboard views from minutes on raw queries to under a second

Python, Node.js, SQL, Google BigQuery, Cloud Functions, Cloud Scheduler, Cloud Run, FastAPI, React, Secret Manager, REST and webhook integrations.