Not another lecture course, but a simulator: you get "hired" by the Datacore company and solve 23 real business problems — from your first EDA report to deploying a model to production and defending a project before the CTO.
Three tracks as one arc: classic ML on business problems, an agent closing an operational process, and a mock certification exam. Plus three real interview questions — the same ones the simulator uses. Pause any scene and read it at your own pace.
We close the gap between "finished a course" and "can do the job": we teach work situations, not algorithms in a vacuum.
You finished Andrew Ng or a bootcamp, but interviews ask about real cases, data leaks and business-driven metrics. Here you'll gain that experience.
You're already on a team but feel the gaps: validation, imbalance, deployment, drift. The simulator is eighteen months of work experience, compressed.
You know Python — what's missing is ML thinking: framing the problem, choosing the metric, and shipping a model to production. That's exactly what we train.
Every module is a work week at Datacore: a marketplace with a fintech product where you've been "hired" as a junior.
Team lead Lena, senior Igor and PM Max set tasks the way it happens in real life: "the warehouse is overstocked", "the model fell apart in production", "marketing argues about discounts".
Theory comes exactly at the task's scope, based on the best courses (Andrew Ng's ML and Deep Learning Specializations) — but translated into the language of work situations, with visual diagrams.
Real Python/sklearn/PyTorch code, typical pitfalls and a "senior's" code review: what's wrong with the split, where the leak hides, why the metric lies.
5–10 test questions with every answer explained. 70%+ passes the module. Complete all 20 core modules — and receive the Middle-track ML Engineer certificate.
From the first day at the office to defending a project before the CTO. The level rises gradually: Junior → Junior+ → Middle-track → Middle → Middle+.
Modules 21–23 are a bonus track based on the book "Machine Learning System Design" (Babushkin, Kravchenko, Manning). Every module includes visual diagrams: from gradient descent to recommendation architectures.
From Junior to Senior AI Engineer. Twenty cases on building agentic assistants inside a real company — and taking them to market. One storyline: the CEO asks for a "marketing factory", you turn the idea into a working system.
The track is part of the same $20/month subscription — no extra cost. The first two modules are free, just like in the main course. The whole track earns a separate "AI Agent Engineer" certificate.
Twenty modules cover all thirty objectives of the official CCAR-F exam, distributed by domain weight. It ends with a mock exam: 60 questions in 120 minutes, a 720 threshold on a 100–1000 scale, and a report giving percent correct per domain — the shape of the real one.
The track is part of the same $20/month subscription — no extra cost. The first two modules are free. Completing the track and passing the mock exam earns a "CCAR-F Exam Ready" certificate.
Independent preparation. This course is not affiliated with Anthropic and is not an official certification; the CCAR-F exam is taken separately.
A separate practice section: answer the question out loud — open the model answer — compare. That's how people prepare for real interviews.
12 screening and first-technical questions: ML basics, metrics, overfitting, validation. Free after registration.
12 production questions: leaks, imbalance, A/B tests, drift, deployment. Everything separating "took a course" from "ran live systems".
8 full design-interview tasks — recommendations, anti-fraud, demand forecasting, search, ETA, scoring, moderation, LLM support — with framework-guided model answers.
Start for free. Cancel the subscription anytime — your progress is kept.
Basic Python — yes (variables, functions, loops). Pandas, sklearn and PyTorch are introduced gradually inside the modules. If Python is new to you, take any one-week intensive in parallel.
Courses teach theory by topic. We teach work situations: "the model fell apart in production — find out why", "marketing demands a rollout — prove it's wrong". The theory inside rests on the classic curricula (Andrew Ng and others), but the structure is a career, not a textbook's table of contents.
A module takes 45 minutes to 3 hours. At a pace of "2 modules a week" the whole path to the certificate takes 10–12 weeks.
There are three, and they are independent. "Middle-track ML Engineer" is issued for the 20 core modules and the final exam. "AI Agent Engineer" is issued for the 20 modules of the agentic engineering track. "CCAR-F Exam Ready" is issued for the 20 modules of the certification track plus a mock exam scored 720 or higher. All are personal, carry a unique ID, and can be printed or added to LinkedIn. Note that the third attests readiness for the Claude Certified Architect exam rather than replacing it — only Anthropic issues the official certification.
Yes, anytime. PRO access remains until the end of the paid period, and your progress and certificate are kept forever.
Registration takes 30 seconds. The first two modules are waiting for you, free.
Start the simulator →