Academy
Course 101Beginner

AI Training Foundations

Start here. Learn the difference between pretraining, fine-tuning and prompting, how to turn a vague idea into a measurable task, how to pick a base model, and why a small, clean dataset beats a large, messy one.

A faceted obsidian stone with a glowing red seam, and a glass sapling growing from its top

Lessons

  1. 1. What it means to train an AI model

    Pretraining, fine-tuning, preference training and prompting are four different levers. Knowing which one you're pulling saves weeks.

    9 min

  2. 2. Define the task and its success criteria

    A model can only get better at something you can measure. Turn an idea into a task spec with inputs, outputs and a score.

    8 min

  3. 3. Choose a base model

    Bigger is usually more capable and always slower and dearer. Pick on measured accuracy against your latency and cost budget.

    7 min

  4. 4. Curate data that teaches

    A thousand carefully chosen examples can do more than a hundred thousand scraped ones. Clean, cover the edge cases, hold some back, and write it down.

    10 min

Sources

Every source cited across this course's lessons.

  1. [1]Vaswani, A. et al. (2017). Attention Is All You Need. NeurIPS 2017 · arXiv:1706.03762.
  2. [2]Kaplan, J. et al. (2020). Scaling Laws for Neural Language Models. arXiv:2001.08361.
  3. [3]Brown, T. et al. (2020). Language Models are Few-Shot Learners. NeurIPS 2020 · arXiv:2005.14165.
  4. [4]Hu, E. et al. (2021). LoRA: Low-Rank Adaptation of Large Language Models. ICLR 2022 · arXiv:2106.09685.
  5. [5]Ouyang, L. et al. (2022). Training language models to follow instructions with human feedback. NeurIPS 2022 · arXiv:2203.02155.
  6. [6]Zinkevich, M. (2017). Rules of Machine Learning: Best Practices for ML Engineering. Google for Developers.
  7. [7]Liang, P. et al. (2022). Holistic Evaluation of Language Models. TMLR 2023 · arXiv:2211.09110.
  8. [8]Vercel (2025). AI Gateway documentation. vercel.com/docs.
  9. [9]Zhou, C. et al. (2023). LIMA: Less Is More for Alignment. NeurIPS 2023 · arXiv:2305.11206.
  10. [10]Gebru, T. et al. (2018). Datasheets for Datasets. Communications of the ACM · arXiv:1803.09010.