CS 145 Projects

Choose your mission: Master SQL, dive deep into data science, or build powerful data systems

At-a-glance: which project is which?

Project	Team	Effort	Focus	Skills you'll build	Prereq
Project 1 BigQuery's First Heist	Solo	~15–20 hrs	Write decision-worthy SQL on a 500 MB+ public dataset	Correlated subqueries, CTEs, window functions, paper debug tables	None
P2 · Science BigQuery Part Deux	Solo or pairs	~30–40 hrs	Enrich with new data, visualize, train a BigQuery ML model	CSV upload & joins, matplotlib, BigQuery ML, query plan analysis	Complete P1
P2 · Systems NanoQuery Mogrifier	Solo or pairs	~30–40 hrs	Build a columnar query engine from first principles	Parquet I/O, HPJ & SMJ from scratch, query planning, benchmarking	Complete P1

Everyone does Project 1. For Project 2, you pick one path: Science or Systems — not both. Both options take roughly the same time (~30–40 hrs) and earn the same credit. Pick by interest, not by difficulty.

Project 1: BigQuery's First Heist

Everyone must complete Project 1, whether you're a minion or a mastermind!

Gru's stolen a massive dataset! Dive into 500MB+ of BigQuery data and extract insights so brilliant, they'll make his shrink ray jealous.

Mission Briefing:

Pick your dataset. Write SQL that sparkles. Present findings that'll make Gru cackle with glee.

Skills Unlocked:

Master SQL subqueries, CTEs, and window functions.

Getting Started:

See Ed #project1 Announcements – Everything you need to get running with BigQuery

View Full Project Requirements →

You need to do this solo.

Credit: Images by GPT4o. Shiva had too much fun prompting for these random scenarios. Especially getting Gru's smirks just right. If you make better ones, lmk.

Project 2: Pick ONE Path

Choose either Data Science or Systems — not both.

Both take ~30–40 hrs and earn equal credit. Pick by interest.

Data Science Track

BigQuery Part Deux – Dig Deep

Project 2 Data Science: BigQuery Part Deux

Gru demands you dig deeper into the data mines…

Surface-level insights? BAH! Gru demands DEEPER! Merge datasets, craft visuals that pop, and build ML models that predict the unpredictable.

Mission Briefing:

Expand datasets. Create jaw-dropping visualizations. Deploy ML models. Impress evil investors.

Skills Unlocked:

Advanced SQL, data visualization, machine learning in BigQuery, and query optimization.

View Full Project Requirements →

Team of two or solo.

Systems Track

The NanoQuery Columnar Mogrifier

Gru demands a columnar database that bends the laws of physics…

Build a columnar database so fast it breaks physics! Generate gigabytes, optimize joins to milliseconds, and benchmark like a supervillain.

Mission Briefing:

Parquet storage. Lightning joins. Compression that crushes. Performance that dominates.

Skills Unlocked:

Parquet-based storage, advanced query optimization, SQL parsing, and performance benchmarking.

View Full Project Requirements →

Team of two or solo.

AI Policy: Learn Like You'll Work

Build skills for the AI-augmented future of engineering

The New Reality

Modern engineering requires mastering both AI tools AND fundamentals. Here's why:

Your future job: You'll ship fast with every AI tool available. But you'll need fundamentals to verify correctness, ensure the right semantics, optimize performance, and ask the right design questions.
CS145 paper tests & whiteboard interviews: No AI mode. We test if you can design solutions and debug problems using pure understanding. This is what separates engineers from copy-pasters.

Use these projects to prepare for both: leverage AI to accelerate your learning, but bring your brain to every problem. Jobs don't need highly trained copy-pasters.

Experiment Freely

Use your favorite LLMs—GPT, Claude, Gemini. Try different prompting strategies. Compare their approaches. This is how modern engineers work at startups and big tech alike. Just credit the AI tools you use in your submissions.

Understand Deeply

Build real understanding of:

Query semantics – what the SQL actually does
Performance debugging – why queries are slow
Index strategies – when they help or hurt
Join algorithms – which one fits your data
System tradeoffs – memory vs. disk, latency vs. throughput

Whiteboard interviews and paper tests assess full-stack understanding. Jobs require shipping with AI. Master both modes here.