Learning Tracks
Your roadmap to real skills.
Many tracks and dozens of sub-tracks. Every track starts with a free intro lesson.
Learning Tracks
Many tracks and dozens of sub-tracks. Every track starts with a free intro lesson.
Adapt real models with LoRA/QLoRA on a student GPU budget.
~11 hours · 8 lessons
View track details →The Adaptation Lifecycle
Map pretraining to SFT to alignment and decide between prompting, retrieval, and fine-tuning.
Full Fine-Tuning vs PEFT
Compare full fine-tuning and PEFT and estimate the GPU memory each needs.
LoRA and QLoRA
Explain low-rank adapters and 4-bit QLoRA, and run a LoRA fine-tune.
Instruction Tuning and Preference Optimization
Explain SFT and preference optimization (DPO and the RLHF it replaced for most people).
Data Curation for Fine-Tuning
Build and decontaminate a fine-tuning dataset and format it with the model's chat template.
Compute Realities for Fine-Tuning
Fit jobs on free/cheap GPUs with mixed precision, gradient checkpointing/accumulation, and tracked, seeded runs.
Passion Project: Fine-Tune a Model
Fine-tune a small open model with LoRA, measure it honestly vs the base model, and publish the adapter + recipe.
Fine-Tuning Job Readiness
Translate fine-tuning skills into applied-scientist / ML-engineer resume bullets and interview answers.