Week 0 · Onboarding
- Light load, about four hours.
- End-to-end environment check, primer pack, Docker refresher, API keys, cohort kickoff.
- No new content: we remove friction before it costs a week.
Accelerator
The Applied AI Engineering Accelerator
13 weeks, online, live. 24 two-hour sessions, two per week. 23 seats.
Apply● Next cohort
21 September 2026
The promise, exactly calibrated
After 13 weeks, graduates can join a software team and independently implement a bounded AI feature, while receiving architectural and code-review support for more complex systems.
That is what thirteen weeks honestly allow. We promise no job, no contract, no income. We measure our outcomes and publish them.
The individual requirement
Every participant ships a feature they own, with their own evaluation score and their own PR history, inside a real team delivery.
That is what separates your portfolio from a toy project built alone: you shipped in a team, for a client, under review.
What you leave with
Your feature deployed in a real client project. An evaluation harness with a measured score. Your contribution history. A written evaluation report. And an attestation, which is only the wrapper around that proof.
The lab
Block III relies on real mandates. We work with external organisations, small businesses, nonprofits and technical teams, and our own product backlog serves as a safety net: if an external client drops out, the team ships against a real need from our lab instead of a fictional exercise.
No Block III project is a toy project.
Structure: week zero, then 4 + 4 + 4
Block I · Foundations, weeks 1 to 4
- W1, model mechanics: tokens, embeddings, context windows, model selection, costs.
- W2, prompts and structured outputs: system prompts, JSON schemas, validated tool calling.
- W3, retrieval foundations: chunking, vector stores, baseline RAG and its six failure modes.
- W4, security and testing from day one: prompt injection, guardrails, secrets, PIPEDA and Law 25 markers.
Block II · Engineering, weeks 5 to 8
- W5, production retrieval: hybrid search, reranking, rank fusion, contextual enrichment.
- W6, evals and tracing: golden sets, LLM-as-judge, harness wired into continuous integration, observability.
- W7, agents: tool loops, LangGraph, MCP, human-in-the-loop, failure recovery.
- W8, deployment and the fine-tuning decision: budgets, cache, streaming, prompt framework, retrieval or fine-tuning.
Block III · Production, weeks 9 to 12
- W9, teams and scoping: teams of three or four, real client mandate, reviewed architecture proposal.
- W10, build sprint 1: standups, peer code review, each participant owns one feature.
- W11, sprint 2 and hardening: regressions green, tracing active, documentation, client check-in.
- Delivery: demo in front of the client, evaluation reports, mock technical interviews.
Keep the programme handy.
The short outline, two pages, in French and English. The thirteen weeks, admission, and what you ship at the end.
Admission
Open to any starting language. Teaching is in JavaScript/TypeScript and Python. Other languages or tools may be used when the project calls for them.
The bar: comfortable in one language, able to build a basic web API, familiar with Git and databases, and able to debug ordinary code without depending entirely on an LLM.
Two parts: a guided exercise of about 20 minutes on your own, then a 45-minute live interview, camera on.
Not required for admission: math, PyTorch, model training, frontend skills, a personal GPU.
And one prerequisite people always forget: time. Reserving eight hours a week predicts success better than any technical skill.
Fees
Fees are shown by zone. The full pricing-zone rule is in the terms.
Included: the 24 live sessions, recordings, all materials and repositories, API keys, the community, the client project and the mock technical interview.
No application fee. Payment after admission. Full refund through the end of week 2.
Mobile payment for Zone A, card and bank transfer elsewhere, employer billing possible. Three sponsored seats are reserved per cohort.
| Zone A | Zone B | Zone C | |
|---|---|---|---|
| Accelerator, 13 weeks | CA$980 (≈ US$700 / €610), or 3 × CA$345 (≈ US$245 / €215) | CA$1,930 (≈ US$1,385 / €1,200), or 3 × CA$675 (≈ US$485 / €420) | CA$2,895 (≈ US$2,075 / €1,800), or 3 × CA$1,045 (≈ US$750 / €650) |
No application fee. Full refund through the end of week 2. Coding with AI graduates pay 17% less, with no deadline. Indicative rounded equivalents.
Which zone am I in?
Your zone follows your country of residence, not your nationality. It is set at admission and does not change afterwards.
- Zone A Cameroon, Senegal, Côte d'Ivoire, Nigeria, Morocco, India
- Zone B South Africa, Gabon, Mauritius, Brazil, Mexico, Türkiye
- Zone C Canada, France, Belgium, Switzerland, United States, Germanyprobably yours
The three zones follow the World Bank country income groups.
Visitor country inferred with the IP to Country Lite database by DB-IP.
See all 218 countries, zone by zoneCalendar
Cohort under way. Final decisions 11 September. Week 0 from 14 to 20 September. Session 1 on 21 September. Demo day in mid-December.
Frequently Asked Questions
Do I need to know JavaScript, TypeScript or Python to apply?
No. Any starting language is accepted. Navigating an unfamiliar codebase is part of the job, and it is one of the things the admission exercise observes.
Will I find a job afterwards?
We do not promise that, and be wary of anyone who does. What we guarantee is different: you leave with a feature deployed in a real project, a measured score and a contribution history. The programme also includes a CV review and a 45-minute mock technical interview. It does not provide placement or employer introductions.
How much time per week, really?
About eight hours: two live two-hour sessions, plus four hours of labs and reading. Block III asks for a little more on some weeks, when the client is waiting for a delivery.
My connection is unstable. Can it work?
Yes. Sessions are recorded and downloadable, materials are in PDF and on GitHub, the community is on WhatsApp, and audio-only is perfectly acceptable during sessions. Just say so; nobody will hold it against you.
How is this different from Coding with AI?
Coding with AI makes AI your tool: the deliverable is an ordinary feature in your own codebase. The Accelerator teaches you to build AI products: the model becomes a component in your system. Neither is a prerequisite for the other, and neither waives the admission interview.
Can my employer fund my seat?
Yes, we issue an invoice to your organisation. Indicate it in your application.
What happens if the programme is not right for me?
You request a full refund through the end of week 2, without having to justify it.
23 seats.
Free application, response within 72 hours.