Report a bug

What went wrong?

Advanced Prompt Engineering — AI Academy

AI Academy/Courses/Advanced Prompt Engineering

⌘K

UK Sign in Sign up

AdvancedИнженерия

Advanced Prompt Engineering

Deep dive into prompt engineering techniques: chain-of-thought, few-shot, self-consistency, tree-of-thoughts, ReAct, meta-prompts, XML structuring, adversarial testing, and automated prompt optimization. Applies to all top LLMs: Claude, GPT-4, Gemini.

5modules

14lessons

390 mintotal time

Developers and AI engineers building reliable LLM pipelinesaudience

Progress

0%0 / 14 lessons

Sign in to save your progress.

Module 1

Foundations of Systematic Prompting

The LLM mental model, prompt structuring (XML/markdown, roles), and balancing zero-shot vs few-shot.

LLM as a text completion engine

Why an LLM is not a database and not an oracle but a probabilistic next-token predictor, and why prompt wording dramatically changes the output.

Structuring prompts: XML, markdown, roles

XML tags for Claude, markdown for GPT/Gemini, separating system/user/assistant, and defending against delimiter injection.

Zero-shot vs few-shot: when and how

When zero-shot is enough, how to pick few-shot examples (diversity, format matching), negative examples, and dynamic selection.

Module 2

Reasoning Techniques

Chain-of-thought and self-consistency, tree-of-thoughts, ReAct — how to make the model reason reliably.

Chain-of-thought and self-consistency

CoT patterns, zero-shot CoT 'Let's think step by step', self-consistency (sampling + voting), when CoT hurts.

Tree-of-Thoughts

The ToT algorithm, BFS/DFS exploration, evaluator prompts, a practical framework-free implementation.

ReAct: reasoning + acting

The ReAct pattern (Reason + Act), tool-augmented reasoning, the thought/action/observation loop, a manual implementation with function calling.

Module 3

Instruction Design Patterns

Roles and personas, output constraints, and meta-prompts — how to precisely control model behavior.

Roles, personas, and system-prompt hardening

Role prompting, persona locking, hardening the system prompt against jailbreak, the difference between role and identity.

Output constraints: format, length, negative

Format constraints (JSON/CSV/XML), length, negative ('do NOT'), constraint ordering (Claude — at start, GPT — at end).

Meta-prompts and self-improving

Prompt generators, self-improving prompts, output self-critique, the constitutional AI pattern.

Module 4

Reliability & Evaluation

Self-consistency in production, adversarial robustness, automated prompt-quality evaluation.

Self-consistency in production

Self-consistency decoding, majority-voting implementation, ensemble prompts, cost vs accuracy tradeoffs.

Adversarial: injections, jailbreak, red-teaming

Prompt injection, jailbreak vectors (DAN, role-play, encoding tricks), defensive patterns, red-teaming your own prompts.

Prompt evaluation: LLM-as-judge and test harnesses

The LLM-as-judge pattern, rubric design, pairwise comparison, automated harnesses (PromptFoo), regression testing of prompts.

Module 5

Optimization & Scale

Automated prompt optimization (DSPy) and production practices: versioning, A/B, cost, drift monitoring.

Automated optimization: DSPy

DSPy overview, automatic few-shot selection, prompt optimization loops, when it beats manual tuning.

Prompts in production: versions, A/B, cost, drift

Prompt versioning, A/B testing, cost optimization (caching, shorter prompts), drift monitoring, prompt as code.

AAI AcademyEN Start learningCatalog

Overview

AI Academy Progress Achievements Leaderboard Insights Notes Notifications Team API Pricing Settings

Learning

Tracks8 Courses51

Libraries

Prompts30 Workflows12 Tools22 Compare Blog

Updated · May 14, 2026