120 short chapters to learn it, then 8 production incidents and 39 design drills to practise it, including real outages from Cloudflare, GitLab and others.
Rather read first? Browse all 120 chapters
Every expert was a beginner who didn't quit. Take it one chapter at a time.
11 pieces in this demo's split. A simplified splitter, not a real tokenizer — real models use a learned vocabulary, so their count will differ. It's the model's count that gets billed and limited, which is why chunking is sized in tokens. Read the chapter
5 tracks, each a short ladder of chapters. Start anywhere — AI is the gentlest opening.
AI Engineering
43 chapters · LLMs, prompts, RAG, agents, multimodal, prod — beginner to practitioner.
Start AI
ML (Beginner)
11 chapters · Classical machine learning with scikit-learn: data → models → shipped.
Start ML
DSA (Beginner)
21 chapters · Patterns, not random problems. Warmup → core patterns → mocks.
Start DSA patterns
System Design
34 chapters · From first principles to production trade-offs: data, caching, messaging, key technologies and full design problems.
Start system design
Something is broken and you're on call. Read the logs, pick the fix, and watch the system come back to life. Every component links to the lesson behind it.
Payments double-charged
During a gateway slowdown, customers are charged twice and some paid orders never ship. Trace it from the retry loop to the warehouse.
Start the incident
The 9-second dashboard
Support agents wait nine seconds for their dashboard. Find four separate performance problems, each hiding behind a plausible red herring.
Start the incident
The AI agent gone wrong
A refund agent loops, forgets the user, emails customers three times and burns six times its budget. Fix seven parts of it.
Start the incident
The failover that lost orders
The database failed over cleanly, or so it seemed. Orders vanished, IDs collided and a bad delete spread everywhere. Fix six weak points.
Start the incident
Black Friday outage
Checkout is down in the middle of a sale. Fix four failures, one hop at a time, and get traffic flowing again.
Start the incident
Bring the store back online
Ten systems behind an online store are down. Fix them in any order, then see what shoppers can actually do.
Start the incident
Kafka consumer lag
Order events are piling up and confirmations are hours late. Trace the lag from producer to consumer and fix four causes.
Start the incident
The HR chatbot gives wrong answers
A RAG chatbot answers HR questions badly. Fix eight parts of its pipeline, then see which answers employees can trust.
Start the incident
Here is a system and how it is configured. The requirements just changed. What do you change? Short drills on Redis, Kafka, RabbitMQ, SQS, Hazelcast, caching and events, from beginner to senior.
The product cache outgrows its memory
A Redis cache with the default eviction policy meets a catalog three times bigger.
Try the drill
A FIFO queue meets Black Friday
Order updates use one message group. Traffic is about to grow ten times.
Try the drill
A slow fraud check joins the consumer
Each record now takes up to 800 ms, and every flagged order must be written exactly once.
Try the drill
Orders span payment, restaurant and courier
A synchronous chain of calls needs manual fixes whenever a step fails halfway.
Try the drill
Pick one and we’ll point you at the friendliest first step.
Twenty lines gets you a first LLM call. Then prompts, then RAG. Start here
Answer one question, finish a chapter or try a drill to start a streak.
📣 Share your streak
Turn your progress into a card.
LLMs, prompts, RAG, agents, multimodal, prod — beginner to practitioner. · 43 short chapters
Welcome
Starts with What is AI Engineering? · 2 lessons
AI Foundations (No Math Fear)
Starts with What is AI, Really? · 4 lessons
Python for AI
Starts with Python Essentials for AI · 3 lessons
How LLMs Work (Visual First)
Starts with What are LLMs? · 5 lessons
Your First AI App
Starts with Making Your First LLM API Call · 3 lessons
Prompt Engineering
Starts with What is Prompt Engineering? · 5 lessons
Pattern-first lessons, runnable code, quizzes and mock interviews — a beginner on-ramp built around common interview patterns and worked examples.
Learn the core patterns — two pointers, sliding window, hash tables, binary search, trees — then rehearse with timed mocks.
Learn to design systems under interview conditions — tradeoffs, estimates and structured walkthroughs starting with TinyURL and a rate limiter.
You'll never learn alone
Quizzes explain themselves, notes remember for you, streaks keep count.
Listen on walks
Every chapter reads aloud, so you can learn with chai instead of stress.
Tiny celebrations
A little confetti on every chapter you finish. Small wins, often.
Minute-level detail on when to use what, with links to the official docs. More showdowns land regularly.
pandas vs Polars vs FireDucks
DataFrames compared: compatibility, speed model, and which to learn first.
Compare
Chroma vs Qdrant vs Pinecone
Vector DBs for RAG: local prototyping vs self-hosted vs fully managed.
Compare
Choosing an LLM API
OpenAI vs Anthropic vs Gemini: how to pick a first provider and stay portable.
Compare
LangChain vs LlamaIndex vs Raw SDK
Frameworks compared: agents and prototypes vs data-first RAG vs full control.
Compare
pgvector vs Dedicated Vector DBs
Postgres extension vs Chroma/Qdrant/Pinecone: one stack or best of breed?
Compare
Jev vs JEPA vs LLM
Typed decisions vs world-model research vs text generation: what each one is, and when it fits.
Compare
Kafka vs RabbitMQ vs SQS
A replayable event log, a flexible message router, or a fully managed queue: which one fits your system.
Compare
RAG vs Fine-Tuning vs Long Context
Three ways to give a model your data: look it up, train it in, or paste it all. When each one wins.
Compare
REST vs gRPC vs GraphQL
Resources over HTTP, fast typed calls between services, or one flexible query endpoint: picking an API style.
Compare
Ollama vs vLLM vs llama.cpp
Running open models yourself: the easiest local setup, the high-throughput server, or the portable engine underneath.
Compare
Temporal vs Airflow vs Step Functions
Durable business workflows, scheduled data pipelines, or AWS-managed state machines.
Compare
Short chapters, quizzes that explain themselves, and a streak that keeps count for you.
Read one cozy chapterPrefer to browse first? See every track.