AI Exercises

Introduction

Practical exercises are the fastest way to internalise everything you've learned. Below is a curated set spanning prompt engineering, RAG, agents and deployment — pick one a week and ship it.

Beginner analogy: Like LeetCode for AI engineers — build muscle by repetition.

Understanding the topic

Core concepts to understand:

• Build a chatbot with streaming + memory.
• Implement RAG over your own notes.
• Build a tool-calling agent that searches the web.
• Add per-user budget caps to an AI feature.
• Set up evals for one prompt + measure accuracy.
• Deploy a multi-step workflow with LangGraph.
• Implement prompt caching and measure cost reduction.
• Build a Perplexity-like research agent.
• Add observability (LangSmith / Helicone).
• Red-team your own prompt for injection.

Syntax reference

Visual workflow / architecture:

Week 1 → chatbot
Week 2 → RAG
Week 3 → agent + tool calling
Week 4 → eval + monitoring
Week 5 → cost + caching
Week 6 → red-team + safety

Real-world use

Many top AI engineers got hired by shipping these exact mini-projects on GitHub.

Best practices

• Ship one project per week, however small.
• Open-source what you build — it's your portfolio.

Hands-on exercise

Interview preparation — practice these questions:

• Q1. Walk me through a project you've built.
• Q2. What was the hardest bug?
• Q3. How did you evaluate it?