Bootcamp Scheduler — AI Development Practical

A hands-on practical where you build a real constraint optimisation system, step by step, using AI.

Setting Up Your Computer

Before starting, you need Python and the project dependencies installed. This is especially important on Windows, where Python setup can be tricky.

Open Claude Code in this directory and give it this prompt:

Check if my computer is set up for this practical. I need Python 3.10 or later. Check what Python versions I have installed. If I don't have one, install one for me. Then create a virtual environment using that Python version, activate it, and install the dependencies from requirements.txt. Run python -m pytest tests/ -x to verify everything is working — the tests should fail with import errors (that's expected, we haven't built anything yet). Tell me when I'm ready to go.

Claude will handle the rest — including downloading Python if needed, dealing with Windows PATH issues, and creating the virtual environment.

If you don't have Claude Code installed yet, run:

npm install -g @anthropic-ai/claude-code

This requires Node.js 18+. See claude.ai/code for details.

---

The Scenario

You are building a scheduling tool for a training company. They run 2-day bootcamps and need to assign trainers from a pool, respecting availability, experience requirements, weekly limits, locations, and bank holidays.

The trainer availability comes in as Excel spreadsheets (in the data/ folder). Your job is to turn that into an optimal schedule.

The Method

This practical teaches a spec-first workflow — you describe what to build, and Claude builds it:

1. Spec it — Write a precise technical specification (you'll build a /spec command for this)
2. Build it — Prompt Claude to implement from your spec
3. Test it — Run the tests to verify it works
4. Repeat — Each challenge adds new requirements

The Challenges

There are 10 challenges, getting progressively harder:

#	Challenge	What You Build
1	Spec Skill	A /spec command that interviews you and writes a technical spec
2	Build Skill	A /build command that implements code from a spec
3	Availability Parser	Read trainer availability from Excel
4	Find Valid Slots	Generate valid 2-day bootcamp time slots
5	Greedy Scheduler	Assign trainers to slots greedily
6	Experience Rules	Ensure each bootcamp has an experienced trainer
7	Constraint Solver	Optimal scheduling using Google OR-Tools CP-SAT
8	Caps and Limits	Enforce per-trainer and weekly caps
9	Location Awareness	Assign bootcamps to locations with language requirements
10	Full System	Bank holidays, weightings, travel penalties — the lot

Challenges 1–2 are about building your own tools. Challenges 3–10 are about using those tools to build the scheduler itself.

Getting Started

Open Claude Code in this directory:

claude

Then type:

/next

This will show you your current progress and tell you exactly what to do next. Use /next whenever you're unsure where you are.

Structure

challenges/     Ten progressively harder challenges
data/           Sample Excel input files (basic, intermediate, advanced)
specs/          Your specifications live here (example provided)
tests/          Tests that verify each challenge
scheduler.py    Your implementation (starts empty)

Tips

• You don't need to be a Python expert — that's what Claude is for
• Talk to Claude in plain English. If something is too technical, ask it to explain more simply
• Every challenge has tests. If the tests pass, you've done it right
• Use /next to check your progress at any time
• If you get stuck on a challenge, read the challenge file in challenges/ — it has all the details