The Build Lab
Six months from your first dataset to an AI system you can defend.
Don’t just learn AI, ship systems that work. Every stage of the program ends with something you can demonstrate, explain and add to your portfolio, and every submission is read by a mentor before it counts as finished.
Ten more you could build next
Once the core five are done, members pick from a longer brief list — or bring a problem from their own job and build against that instead.
How work is assessed
Your work won’t just be graded.
It will be challenged.
Every project is submitted, read line by line and returned with written comments. You fix it and resubmit. That loop — not the lecture — is where the improvement happens.
- Line-level comments within three working days
- A rubric covering correctness, reasoning and communication
- At least one resubmission expected on every project
- A final portfolio review before you start applying
Mentor review
- ✕Target leaked into a feature via a post-event column
- ✕Accuracy reported on an imbalanced set, no baseline
- ✕One 400-line notebook, no functions, no seed
- ✕Nothing runs outside the author’s machine
First submission. Every point here was raised by a mentor, in writing.
Screens
What the finished work looks like
Interfaces from recent cohorts. Every project ships with a live demo, a public repository, an architecture write-up and a mentor’s review attached.
01Data Intelligence Dashboard
Rubric score 86Transform messy datasets into a stakeholder-ready SQL and Power BI dashboard that answers a question somebody actually asked.
02Production Prediction API
Rubric score 91Train, evaluate and deploy a machine-learning model behind a working API — with logging, versioning and a threshold you can defend.
03Customer Segmentation Engine
Rubric score 88Use clustering and dimensionality reduction to find behavioural groups a marketing team can name, defend and act on.
04RAG Knowledge Assistant
Faithfulness 84Build an assistant that retrieves information, cites its sources, refuses when it should, and can be evaluated on a fixed question set.
05Agentic Workflow Capstone
Capstone score 89Your own end-to-end system: a model in a loop with tools, memory and guardrails, deployed and measured against a real task.
Interface mock-ups — replace with real learner screenshots and demo recordings before launch
JOIN NOW
Open your first project brief
Module I opens with a dataset, a question and a deadline. You can start it today.
Join any time · 6,000 learners so far