Skip to main content

Intro to AI Engineering

Placeholder: Banner Image for Intro to AI Engineering - Neural Networks and Code

Goal​

Master the tools and habits that separate "model demos" from production AI.

You will build deep learning intuition (how training works and why it fails), gain practical LLM mental models, and develop PyTorch fluency. Alongside that, you'll practice clean coding so your ML codebases stay testable, debuggable, and easy to extend.

By the end, you can:

  • Implement and train neural networks in PyTorch without copy-pasting templates.
  • Debug common training failures (overfitting, unstable loss, poor evaluation, data issues).
  • Explain the main moving parts of transformers and LLMs at a practical level.
  • Write ML code that stays readable as the project grows.

Estimated Time Impact

38-53 hours total

Deep Learning & LLMs(12h)
PyTorch Engineering(8h)
Engineering Excellence(25h)

1. Deep Learning & LLMs​

Learn the foundations behind modern deep learning, then build a practical mental model of transformers and LLM behavior.

course
advanced

CS231n Winter 2016 (Lectures #2, #3, #4)

6-8 hours
talk
intermediate

A Deep Dive into LLMs

1.5-2 hours
video
beginner

How I Use LLMs (Karpathy)

1-1.5 hours

What You'll Learn

  • Backprop and optimization: what training is actually doing.
  • Why models fail: underfitting, overfitting, instability, and bad data.
  • Transformer basics: tokens, attention, context, and next-token prediction.
  • How to reason about LLMs in practice: training vs inference vs fine-tuning.

2. PyTorch Engineering​

Turn theory into implementation. Build real PyTorch fluency so you can write and debug training code confidently.

tutorial
intermediate

UvA Deep Learning Tutorials: Chapter #2: Intro to PyTorch

4-6 hours

What You'll Learn

  • Tensors, shapes, broadcasting, and device placement.
  • Autograd intuition and how gradients flow.
  • nn.Module structure and clean training/evaluation loops.
  • Debugging patterns: shapes, metrics, overfitting checks, and sanity tests.

3. Engineering Excellence​

Write ML code like an engineer: clean structure, readable logic, and maintainability as experiments scale.

book
intermediate

Clean Code (Ch. 1–10)

15-20 hours
practice
intermediate

LeetCode: Programming Skills Study Plan

10-15 hours

What You'll Learn- Naming and function design that keeps ML codebases understandable.

  • How to avoid "research spaghetti" as experiments multiply.
  • Building implementation fluency so you can translate ideas into correct code faster.