Not a sample, not a first chapter. All 47 modules, all 353 lessons, roughly 438 hours — from what a derivative is through to writing a fused Triton kernel. No card, no trial, no expiry.
Every module ends in a challenge where the concept is the game mechanic. Broadcast shapes to score. Hunt a NaN through a computation graph. You cannot win by guessing.
One ordered path from vectors and derivatives to RLHF, distributed training and inference optimization. No gap where you are expected to already know something.
Every track names what its material is tested on, and the last track is the loop itself — breadth, coding, system design, depth, behavioral, negotiation.
The tutorials are the course. Concepts are introduced in a few minutes and then learned by beating something that will not let you fake it.
In Shape Golf you fix twenty broken forward passes in as few edits as you can. In Reward Hack you are the policy, trying to game a reward model — then you patch the reward model against your own exploit.
Every challenge runs in the browser. No GPU, no environment to set up, no account needed to start — the setup tax that ends most self-taught attempts is not here.
The other half is a question bank of what is really asked, resumes tailored to each posting and applied for you — never submitted without your approval — a tracker built for ML loops, a calendar that blocks out prep, and members inside the companies you are interviewing with. That is Pro: $9,900, one time, 5 years of access.
Forty-seven interactive challenges where the concept is the game mechanic. Broadcast shapes to score. Hunt a NaN through a computation graph. Negotiate against a simulated hiring manager.
Free · In development
Questions actually asked in ML engineering loops, each with a full worked solution, the follow-ups that came after it, and what the interviewer was grading.
Pro · In development
It finds roles on company career sites, writes a resume tailored to each posting, fills in the application on that company's own site, and then stops and waits for you to approve it. Nothing is ever submitted automatically.
Pro · In development
A pipeline built for ML loops specifically — recruiter screen, coding, ML breadth, system design, depth, behavioral, team match — so you always know what is next and what it will test.
Pro · In development
Interviews land in your calendar with the prep blocked out before them, the time zone resolved correctly, and a reminder of what that specific round tests.
Pro · In development
Members working at the companies you are interviewing with — for referrals, for mock interviews, and for the honest answer about what a team is actually like before you sign.
Pro · In development
