The Full Curriculum — SustainOS AI Lab

Week 0

Ready before Day 1

Short pre-recorded lessons, done in your own time. You meet Claude properly, install your first skill, and learn how to brief a build so it comes back right. You arrive on Day 1 ready to build, not ready to set up.

Claude Code installed and running on your own machine
What a skill is, and how to install one in three steps
What a product spec is, and why it decides whether the build works
Why the first version of anything stays deliberately small
The vocabulary you will use for the next six weeks

By Day 1: a working setup and a list of tool ideas from your own job.

Skills awardedClaudeSkillsProduct specsMVP scoping

Week 1

Build your first tool

By Friday you have a working tool, live on the internet, on its own URL. You meet Claude Code, turn a rough idea into a brief precise enough to build from, and put the result online. Not a prototype on your laptop. Something you can send to a colleague.

Turning a vague idea into a precise brief, guided by an interview
Building your first working interface with Claude Code
Publishing it to a live URL anyone can open
Choosing what your tool actually needs, and what it does not
We build a double materiality matrix together, from a complete assessment

By Friday: your own tool, live on its own URL.

Skills awardedClaude CodeFrontendTool deployment

Weeks 2–4 — one tool, growing

Week 2

Build your second tool, and upgrade your setup

A second tool, and a setup that stops slowing you down. The manual steps from Week 1 disappear. Your work saves its own history, publishes itself, and remembers the rules you set, so you spend the time on your thinking instead of on the tooling.

GitHub, so your work has a history and a bad build is one click back
A publishing pipeline, so saving your work puts it live automatically
The documents that give Claude your rules, so it builds your way every time
Connecting Claude to the tools you already use
We start a supplier engagement portal, and grow it for the next three weeks

By Friday: your own tool, in your own repository, publishing itself.

Skills awardedVersion controlAuto-deployGovernanceMCPs

Week 3

Your tool gets a memory

Right now your tool collects, then forgets. This week it remembers. It gets a database, so what people enter is still there tomorrow, structured, in one place you own. And the moment your data lives somewhere you control, everything downstream opens up. Analyse it, refine it, report on it, build a dashboard on top of it. This is the week that makes all of it possible.

What a backend and a database actually are, and how they talk to each other
Setting up your own database and designing what it stores
Deciding what deserves to be a record, and what belongs in a file
Reading your own data, and trusting what you see
Why stored data is the real unlock: analysis, reporting and dashboards all become possible from here, and we build both in Weeks 5 and 6
Handling keys safely, so nothing sensitive ever ends up somewhere public

By Friday: enter something, close the tab, come back tomorrow. It is still there, in your database.

Skills awardedBackendDatabasesData modellingKeys and secrets

Week 4

Protect your tool

Now you decide who gets in. Your tool gets a real login, so only the people you invite can use it, and every entry is tied to someone you can name. This is where a side project becomes something you can put in front of your boss, your IT team, or a client.

How modern logins actually work, and why the best ones use no password at all
Invite-only access, so an address you never invited gets nothing
Signing in with Google, and when that is the better choice
Connecting your tool to outside services, so it can send real email
Putting your tool on your own web address instead of a platform subdomain

By Friday: you sign in through your own link. Someone you did not invite gets nothing.

Skills awardedAuthenticationAPIsEmail serviceDomain and DNS

Week 5

Roles, admins and users

Different people, different views. You build a carbon inventory platform where a plant manager submits their own site’s data and an ESG lead reviews and approves it. Same login, completely different screens, and nobody sees what they should not.

Why getting in and being allowed to see are two different problems
Building roles, so each person lands on the view that belongs to them
Locking data down at the record level, not just at the door
Submit, review, approve, with a named person on every step
Putting the calculation inside the tool, so the person entering data cannot get the maths wrong

By Friday: your tool with roles. The data one role is not allowed to see is gone from their screen.

Skills awardedAuthorizationRolesRow-level securityReview workflows

Week 6

Everything connects

Your capstone. You build a decarbonization roadmap, and then make it read what your emissions platform produces. One database, one login, two tools working as one system. Then we go a level up, so you can place whatever launches next without starting over.

Connecting two tools so one reads what the other produces
One login across everything you have built
What an AI model can and cannot do, in plain terms
RAG: how AI answers from your documents, with references you can check
The difference between a workflow and an agent, and when each one is right
How connectors let your tools reach the rest of your stack

By Friday: your own system, presented to the room.

Skills awardedSystem integrationRAGAgentsAI workflows

Lifetime

The building doesn’t stop

The cohort ends. Your access doesn’t. Every recording, all the material, and a community of sustainability professionals building, breaking, fixing and launching together.

Every recording and all the material, yours to keep
The community stays open, and the sessions keep going
Bring whatever you decide to build next

Your skill set27 of 27 · yours to keep