Deliver a workshop automation assistant.
Combine intake, simulated AI, review, recovery, and evaluation in one explainable workflow.
Can another person run, test, and understand the limits of your complete automation?
Design a fictional workshop assistant that processes enquiries into reviewed drafts. Combine validated intake, simulated model responses, restricted destinations, approval, and an operation ledger. Produce a working local notebook and a pilot plan. A live model or external action is an optional future integration, not a course requirement.
Build it, step by step
- 01
Write the workflow map, contracts, owners, and limits before implementation.
- 02
Load the batch and validate each record, preserving a result per row.
- 03
Deduplicate events, validate simulated proposals, and route uncertain cases to review.
- 04
Apply a versioned approval gate and an operation ledger before a local ready state.
- 05
Inject malformed data, hostile text, duplicate events, stale approvals, and simulated service failures.
- 06
Report test results, review coverage, estimated cost, and limitations. Prepare a rollback and handover plan.
Your project checklist
- Every input is accounted for and no invalid record reaches ready.
- Model output cannot approve itself or choose an arbitrary destination.
- Duplicate, pending, and completed operations have distinct handling.
- Tests cover normal, empty, malformed, repeated, and failure cases.
- The handover includes reproducible fixtures, source versions, owners, budget, and rollback.
- Simulated results are clearly labelled and live integration gaps remain visible.
Free, open and run by a nonprofit.
This page is part of AI Automation, a free course of 20 modules and 60 lessons in the engineering.fit( ) Learning Lab. The Learning Lab is built and run by engineering.fit( ), a nonprofit founded by Abdullateef Opeyemi Bakare that helps African engineering students gain research experience. You do not need an account or pay a fee to learn.