Public deck

AI & Prompt Engineering

Vendor-neutral fundamentals of working with large language models: tokens and sampling, prompting patterns, embeddings and vector search, retrieval-augmented generation, and the failure modes that bite in production.

Engineering

AShared by Alex Chen

0 Play Sign in to clone

You’re reading a free preview. Sign up to see every topic and question — and to play this deck as games.

See the whole deck

Suggested order

Arrows point from what to learn first. Topics side by side can be studied in any order — and cloning the deck brings this order with it.

Topics

Open a topic to read through its content, then play just that topic.

LLM Basics: Tokens, Context and Sampling

What a model actually reads, how much of it fits, and the knobs that decide what comes out.

The prompt shapes that reliably change output quality — and the ones that only look like they do.

Embeddings and Vector Search

Turning text into vectors, measuring closeness honestly, and searching a corpus without comparing against all of it.

Locked · 42 questions inside

Retrieval-Augmented Generation

Putting the right passage in front of the model, and making the answer stay inside it.

Locked · 43 questions inside

Failure Modes: Hallucination, Prompt Injection and Context Rot

How LLM systems go wrong in production, why instructions alone cannot fix it, and what actually helps.

Locked · 43 questions inside

Tool Use and Agent Loops

How a model calls functions, how the act-observe loop is structured, and where the real engineering lives: tool design, stopping conditions, and enforcing permissions outside the prompt.

Locked · 45 questions inside

Evaluation: Benchmarks, LLM-as-Judge and Regression Suites

How to tell whether a change actually helped: golden sets, deterministic checks, model graders and their biases, and why public benchmark scores predict very little about your own system.

Locked · 45 questions inside

Fine-Tuning, Prompting and Retrieval: Choosing Between Them

Which problem each technique actually solves, the order to try them in, and the costs — especially the lock-in — that only show up later.

Locked · 43 questions inside

Cost, Latency and Caching in Production

What a request actually costs and where the time goes: prefill and decode, prompt caching and how to lay a prompt out for it, batching, routing, and measuring the tail.

Locked · 43 questions inside

7 more topics in this deck

Create a free account to open every topic — and play the whole deck without copying it.