Serious ML engineering prep, unlocked.
Monthly
Billed monthly. Cancel anytime.
Our monthly plan grants full access to all premium features. Best for short-term prep before an upcoming interview.
Annual
Billed annually โ $109/yr
Our most popular plan. Previously sold for $28/month โ now only $9/month billed yearly.
Save over 67% compared to monthly.
Your payment is processed securely by Stripe โ we never store your card details.
All problems unlocked
Access every problem across numpy fundamentals, transformer internals, LLM inference, distributed training, and PyTorch โ 100+ problems and growing.
Unlimited code runs
Free users get 3 runs per day. Premium removes all limits โ run as many times as you need without interruption.
Hidden test cases revealed
See the full suite of hidden edge-case tests before you submit. Understand exactly what your solution needs to handle.
Reference solutions
Every problem comes with a clean, well-commented reference implementation. Read it after solving โ or when you're stuck โ to learn the canonical approach.
Mathematical formulations
Full LaTeX derivations for every algorithm. Understand the math behind softmax, attention, Adam, and every other primitive you'll be asked about.
Progress tracking
See which problems you've solved, your pass rate, and streaks. Know exactly where you stand heading into interviews.
No ads
A clean, distraction-free environment while you practice. No banners, no upsells.
Dark mode
Easy on the eyes for long practice sessions.
โTorchedUp is the only platform where I could actually practice flash attention and ZeRO from scratch. Got the offer at a top lab.โ
Anonymous
โThe reference solutions alone are worth it. Seeing the canonical numpy implementation side-by-side with my attempt is how I actually learned.โ
Anonymous
Cancel anytime ยท Secure checkout via Stripe