◆ Cohort 02 · Syllabus

Six weeks. One artefact. Hands on every session.

The full week-by-week of the Academy Cohort — modules, learning outcomes, hands-on assignments, what you ship by the end, and what we deliberately skip. Read it before you apply. We'd rather you walk away than show up expecting the slide-deck version.

Apply for Cohort 02 →Book a Discovery Call
The frame

What you're signing up for.

Three commitments we hold ourselves to before week one starts.

What it is

Six weeks. One artefact. Hands on every session.

90 minutes live each week, plus three to four hours of self-paced work. By week six you ship a real artefact built on your own data — not a slide deck, not a prototype that breaks the moment you show it.

Who it's for

Managers and senior ICs who'll touch the keyboard.

Not for boards (that's the Primer) and not for whole engineering teams (that's Embedded). For the person inside an org who has to make AI-on-real-work happen and currently has no working playbook for it.

What "lands" means

Capability transferred, measured, and supported.

Every week ends with a hands-on assignment scored against an eval. The capstone is signed off by a founder. The 30-day post-cohort window catches what wobbles after you ship.

Week by week

The full curriculum.

Each week: a focused topic, four to five hands-on outcomes, and an assignment that touches your real work. The capstone you build in week six rides on the artefacts from weeks two through five.

Week 1
01

Where AI fits, and where it doesn't.

The honest version of "AI 101." Map the 2026 capability frontier against the work your team actually does, then cut the hype back to a working mental model. By the end you can tell a teammate when AI is the wrong tool — and mean it.

Outcomes
  • Name the four capability classes you'll use this cohort.
  • Identify three workflows in your org with high AI-fit and three with low.
  • Draw a one-page capability map of your team's current stack.
  • Call BS on at least one vendor pitch.
Assignment

Walk one of your own workflows end-to-end and score it against the AI-fit rubric we ship in week one.

Week 2
02

Prompts, but for grown-ups.

Prompting is a software engineering discipline now, not a vibes exercise. We work in a real prompt IDE, build an eval set, and ship two small workflows: a structured-output extractor and a multi-step task runner.

Outcomes
  • Write prompts with versioned inputs, schemas, and evals.
  • Build a five-case eval set and run it against three models.
  • Ship one extractor that returns clean JSON 95% of the time.
  • Measure latency and cost per call.
Assignment

Replace one repetitive document task in your team's week with a prompted workflow that has a passing eval set.

Week 3
03

RAG over your data, without the cargo cult.

Retrieval is where most internal AI falls over. We build a working RAG bot on a slice of your own documents — chunking, embeddings, retrieval, citations — and stress-test it until it breaks. Then we fix it.

Outcomes
  • Ingest a real document corpus and configure chunking/overlap deliberately.
  • Ship a RAG bot that cites sources.
  • Write a retrieval-quality eval and read the failure modes.
  • Decide when to go RAG vs fine-tune vs neither.
Assignment

Stand up a RAG bot on 50–500 of your own internal documents and run a 10-question eval against it.

Week 4
04

Agents and tools without the hand-waving.

Agents are tools-plus-loops, not magic. We build a tool-calling agent that touches one real system in your stack (calendar, tickets, CRM, files), wire in a guardrail, and watch it fail in interesting ways.

Outcomes
  • Define tools as schemas, not vibes.
  • Ship an agent with at least two tools and one guardrail.
  • Instrument runs with traces.
  • Draw the line between "agent" and "scripted workflow with one LLM call" — and know which one your problem needs.
Assignment

Build a single-purpose agent that automates a recurring task in your week and ship it to a teammate to break.

Week 5
05

Process redesign — the actual lever.

The point most cohorts miss. AI doesn't fix process drift — process redesign does, and AI plugs into the redesigned shape. We map a workflow before-and-after, find where the gain leaks, and redraw the org around the automated step.

Outcomes
  • Draw a before-after process diagram for one workflow.
  • Locate the three most common gain-leak patterns in your draft.
  • Write a change-management brief for the affected team.
  • Defend the redesign to a sceptical stakeholder in a 5-minute walkthrough.
Assignment

Pick the workflow underneath your week-4 agent and redraw the process so the gain doesn't leak when you ship it.

Week 6
06

Capstone + roadmap-forward.

You present the artefact you've been building since week one to the cohort and one of our founders. We critique it, name what's production-ready, and map a 90-day roadmap for the next three workflows.

Outcomes
  • Present a working capstone artefact built on real internal data.
  • Receive structured critique from peers and founders.
  • Leave with a 90-day, three-workflow roadmap.
  • Know which artefact moves to production and which gets archived.
Assignment

Demo the capstone, accept the critique, and commit to one production-bound workflow inside your org for Q+1.

Walk-out artefacts

What you ship by week six.

Concrete deliverables built on your own data — portable to your org, reviewable by your boss, runnable on Monday morning.

01

A working RAG bot grounded on 50–500 of your own internal documents, with citations and an eval set.

02

A tool-calling agent that touches one real system in your stack, with traces and a guardrail.

03

A before-after process diagram for one of your team's workflows, with gain-leak annotations.

04

A versioned prompt library with eval sets for at least three workflows, portable to your org.

05

A 90-day, three-workflow roadmap signed off in your week-6 capstone session.

The negative space

What we deliberately skip.

Six weeks is six weeks. We pick what lands and refuse what doesn't — it's how we keep the cohort honest.

Foundational ML theory

No backprop, no transformer internals. If you want the maths, we'll point you to the canonical courses.

Vendor tours and tool comparisons

We pick a working stack on day one and build with it. You can swap providers later — the patterns transfer.

Fine-tuning your own foundation models

It's almost never the right answer for the workflows this cohort targets. We cover when it would be, and stop there.

Logistics

The shape of the six weeks.

The schedule, load, and support window — written so you can hand this to your manager without a back-and-forth.

Format
Live + self-paced · 6 weeks
Live sessions
90 min weekly · Tuesdays 18:00 AEST
Self-paced load
~3 hrs/week, hands-on
Cohort size
12 seats, hard cap
Cohort Slack
Active for the duration
Post-window
30 days · office hours + iteration
FAQ

The questions buyers actually ask.

If yours isn't here, ask it on the discovery call — we'd rather answer in plain English than write more copy.

What's the prerequisite?

Comfort writing English, opening a terminal, and reading a JSON object. No prior ML experience. If you can ship a Notion page and a spreadsheet formula, you're in.

What's the real time commitment per week?

90 minutes live, plus three to four hours of hands-on work. Less if you're fast, more if you're going deep on the capstone. We assume you have a day job.

What's the refund policy?

Full refund up to 48 hours before week one. After that, we credit 50% to a future cohort or a same-quarter Practice engagement. We don't pretend cohorts are risk-free — we keep the price honest instead.

Can my whole team do this in-house instead?

Yes — that's the Embedded format. Four or more seats from one org gets volume rates here, but a 10+ team usually wants Embedded so we can build around your stack and your data. Ask on the call.

What's the cohort size?

Twelve seats, hard cap. Past 12, the live sessions stop being live. We've held this line through every cohort.

Is there a certificate?

Yes, on capstone completion. We don't pretend it's accredited — it's a signal that you shipped a working artefact, signed off by the founders. Most graduates show the artefact, not the certificate.

Cohort 02 · 8 seats left

Apply now or book a 30-minute call first.

Starts 2026-07-14. Applications close 2026-07-04. $1,950 AUD per seat · volume rates from 4 seats.

Apply for Cohort 02
Book a Discovery Call
← Back to Academy