Clone and learn
git clone https://github.com/urmzd/openlearncd openlearnpractice/bin/ol learnSupersource
A free, self-paced curriculum from undergraduate foundations to Staff-level depth. Math first, then the code that implements it, in C, CUDA, Rust, Go, and Python.
Follow a path
Section titled “Follow a path”A path is a reading order for one job, cut across the tracks, with a checkable “done when” for every stage.
Superstar FDEForward deployed engineer at an inference cloud, end to end: all six parts plus a capstone.
LLM FoundationsFrom n-grams to reasoning models: history, transformer math, architecture variants.
Inference PerformanceEngine internals, quantization, the framework map, serving and load testing.
TrainingRL foundations, pretraining, SFT, preference and RL post-training, LoRA.
Frameworks and ModelsPyTorch, JAX, and Keras 3 side by side, then loading checkpoints from the Hub.
AI Full StackToken streaming, retrieval, evaluation, and routing across models.
Field EngineeringDiscovery, sizing, performance testing, POCs, migrations, escalations.
Or go deep on a track
Section titled “Or go deep on a track”Run it, don’t just read it
Section titled “Run it, don’t just read it”Practice
One CLI, three exercise kinds: predict, build, reattempt. Attempts go to a gitignored scratchpad. Start practicing.
Case studies
Real builds generalized into runnable, dependency-free implementations that test themselves. See the builds.