BigQuery — Portfolio Projects
Build these projects to demonstrate real skills to employers. Each project is designed to be interview-worthy — something you can walk through in detail.
Project 1: Partitioned/Clustered Analytics Warehouse
Level: Beginner | Time: 2 days
Load a public or synthetic event dataset (e.g. clickstream, orders, or one of BigQuery's public datasets) into a properly partitioned and clustered table, then write a set of analytical queries against it and measure the bytes-scanned difference against an un-partitioned copy of the same data.
Steps
bq load
PARTITION BY DATE(...), CLUSTER BY on your most commonly filtered columns
INSERT ... SELECT
INFORMATION_SCHEMA.JOBS_BY_PROJECT)
Skills Demonstrated
GitHub Repo Name
bigquery-partitioned-warehouse
---
Project 2: BigQuery ML Churn/Classification Model
Level: Intermediate | Time: 3 days
Build an end-to-end BigQuery ML pipeline entirely in SQL — no data ever leaves BigQuery. Train a classification model, evaluate it, and serve predictions back into a table a dashboard could read from.
Steps
data_split_method or a manual WHERE split)
CREATE MODEL (logistic regression is a reasonable first choice)
ML.EVALUATE and record precision/recall/AUC in your README
ML.PREDICT into a results table
Skills Demonstrated
GitHub Repo Name
bigquery-ml-classifier
---
Project 3: Cost-Governed Multi-Team Warehouse
Level: Advanced | Time: 5 days
Design a BigQuery setup for multiple teams sharing one project (or a few projects) — with cost control, access isolation, and monitoring, the way a real data platform team would run it.
Steps
INFORMATION_SCHEMA.JOBS_BY_PROJECT and turn it into a scheduled report or dashboard
Skills Demonstrated
GitHub Repo Name
bigquery-cost-governed-platform
---
Tips for Great Projects
Make it real. Solve an actual problem, even a small one. "Built a partitioned warehouse for my own Strava/finance/reading data and found a real cost-scanning bug" is more impressive than a tutorial clone.
Document everything. A repo with a great README beats one with better code but no explanation. Include: what it does, why you built it, how to run it, what you learned.
Show your thinking. In interviews, you'll be asked: "Why did you choose X over Y?" Have a reason. Architecture decisions matter.
Iterate publicly. Make commits regularly. Employers look at commit history. 10 commits over a week shows real work; 1 commit with everything shows you copied it.