Free · No account · In your browser

Python + AI: Zero to AI Engineer

Go from zero to an AI engineer by writing real code in your browser: Python fundamentals, then production-grade AI -> prompting, RAG, evals, agents, guardrails, and real portfolio projects. Free, no account.

31 modules 239 lessons 235 auto-graded exercises Self-paced · ~46h of material Real Python, graded by real tests Verifiable certificate
Start the course Verify a certificate
Your progress 0 / 235 exercises complete
Daily goal
Module 1

Python Foundations

Run real Python in the browser and build confidence with values, math, strings, booleans, and reading errors.

✓ Hello, Python exercise 8 min ✓ Variables exercise 9 min ✓ Numbers & Math exercise 12 min ✓ Working with Strings exercise 13 min ✓ True, False, and Comparisons exercise 14 min ✓ Input & Output exercise 10 min ✓ Errors & Tracebacks exercise 12 min ✓ Mini-Project: Tip Splitter exercise 15 min ✓ Concept Check: Python Basics exercise 5 min
Module 2

Control Flow

Branch and repeat logic fluently, with heavy scaffolding at loops.

✓ Branching with if / elif / else exercise 9 min ✓ Boolean Logic: and / or / not exercise 9 min ✓ while Loops: Repeat Until It Changes exercise 12 min ✓ for Loops and range() exercise 12 min ✓ Watch: a Loop Adding Up watch 5 min ✓ break, continue, and the loop else exercise 11 min ✓ Multi-way Branching with match / case exercise 10 min ✓ Mini-Project: Guessing Game Engine exercise 14 min ✓ Concept Check: Control Flow exercise 5 min
Module 3

Functions

Author reusable functions with full signatures, docstrings, type hints, and return-vs-print discipline.

✓ Defining a Function exercise 9 min ✓ Parameters: Passing Data In exercise 10 min ✓ Return vs Print exercise 11 min ✓ Default and Keyword Arguments exercise 11 min ✓ Flexible Arguments: *args and **kwargs exercise 12 min ✓ Scope and the Call Stack exercise 11 min ✓ Docstrings and Type Hints exercise 11 min ✓ Mini-Project: Unit Converter exercise 14 min ✓ Concept Check: Functions exercise 5 min
Module 4

Data Structures

Fluently use list, dict, set and tuple, choose the right one, and write comprehensions.

✓ Lists: Create, Index, Mutate exercise 10 min ✓ List Methods and Slicing exercise 12 min ✓ Tuples and Unpacking exercise 10 min ✓ Dictionaries: Keys and Values exercise 11 min ✓ Dict Methods and Iterating exercise 12 min ✓ Sets: Uniqueness and Membership exercise 10 min ✓ Nesting and Choosing the Right Structure exercise 12 min ✓ Comprehensions exercise 13 min ✓ Project: Contact Book exercise 15 min ✓ Concept Check: Data Structures exercise 5 min
Module 5

Files, Errors & Modules

Read and write text, CSV and JSON, handle exceptions properly, and structure code into modules.

✓ Reading and Writing Text Files exercise 10 min ✓ Working with CSV Files exercise 12 min ✓ Saving Data as JSON exercise 10 min ✓ Handling Errors with try/except exercise 12 min ✓ Raising and Defining Exceptions exercise 12 min ✓ A Tour of the Standard Library exercise 13 min ✓ Modules, Imports and the Main Guard exercise 11 min ✓ Packages, pip and Virtual Environments exercise 9 min ✓ Concept Check: Files, Errors & Modules exercise 5 min
Module 6

Object-Oriented Python

Design classes with __init__, methods, dunder methods, inheritance, @property and @dataclass.

✓ Classes and Objects exercise 9 min ✓ Setting Up With __init__ exercise 11 min ✓ Methods and self exercise 11 min ✓ Dunder Methods exercise 13 min ✓ Inheritance and super() exercise 12 min ✓ Computed Attributes With @property exercise 12 min ✓ Less Boilerplate With @dataclass exercise 11 min ✓ Project: A Safe Bank Account exercise 15 min ✓ Concept Check: Object-Oriented Python exercise 5 min
Module 7

Pythonic Code, Testing & Capstone

Write idiomatic, tested code (generators, decorators, pytest) and ship a multi-file-style CLI. Exit = mid-tier Python.

✓ Idiomatic Style (PEP 8) exercise 9 min ✓ Type Hints & Typing exercise 12 min ✓ enumerate, zip, and sorted(key=...) exercise 11 min ✓ EAFP, Truthiness, and the Mutable Default Trap exercise 13 min ✓ Generators and Lazy Sequences exercise 12 min ✓ Decorators exercise 13 min ✓ Testing with pytest-style asserts exercise 11 min ✓ Debugging: Tracebacks and a Real Fix exercise 12 min ✓ Capstone: Expense Tracker (data layer) exercise 16 min ✓ Capstone: the ExpenseApp exercise 22 min ✓ Concept Check: Pythonic Code, Testing & Capstone exercise 5 min
Module 8

SQL & Databases

Query and shape real data with SQL on a live SQLite database in your browser: SELECT, filtering and sorting, aggregates and GROUP BY, JOINs across tables, safe parameterized writes, and subqueries.

✓ SELECT: Asking a Database Questions exercise 13 min ✓ WHERE, ORDER BY & LIMIT exercise 13 min ✓ Aggregates & GROUP BY exercise 14 min ✓ Joining Tables exercise 15 min ✓ INSERT, UPDATE & DELETE (safely) exercise 14 min ✓ Subqueries exercise 13 min ✓ Concept Check: SQL & Databases exercise 5 min
Module 9

Data Foundations: numpy & pandas

Manipulate arrays and clean, filter, group and merge tabular data in the browser.

✓ Why Data Libraries: Lists to Arrays exercise 10 min ✓ Creating and Indexing Arrays exercise 12 min ✓ Vectorization and Broadcasting exercise 12 min ✓ pandas: Series and DataFrame exercise 12 min ✓ Selecting, Filtering and Sorting exercise 13 min ✓ Cleaning Messy Data exercise 14 min ✓ Grouping and Merging exercise 14 min ✓ Project: Analyze a CSV exercise 16 min ✓ Concept Check: Data Foundations: numpy & pandas exercise 5 min
Module 10

Your First Machine Learning Models

Train and evaluate scikit-learn models and diagnose under/overfitting.

✓ The Machine Learning Mental Model exercise 10 min ✓ Looking at Your Data with matplotlib exercise 11 min ✓ Train/Test Split: Honest Measurement exercise 11 min ✓ Your First Model: a Decision Tree exercise 12 min ✓ Logistic Regression & Probabilities exercise 13 min ✓ Watch: Softmax watch 4 min ✓ Evaluating Models: Accuracy and Error exercise 12 min ✓ Beyond Accuracy: Precision, Recall & F1 exercise 14 min ✓ Regression on Noisy Data exercise 14 min ✓ Watch: Overfitting vs Generalizing watch 5 min ✓ Underfitting and Overfitting exercise 14 min ✓ Random Forests: Wisdom of the Crowd exercise 13 min ✓ Project: Classify Flowers End to End exercise 15 min ✓ Concept Check: Your First Machine Learning Models exercise 5 min
Module 11

Neural-Net Intuition, LLMs & AI Capstone

Build honest neural-net intuition in numpy, reason clearly about how LLMs work, and ship an on-device AI capstone that turns data into a plain-English report.

✓ A Neuron from Scratch exercise 10 min ✓ A Two-Layer Forward Pass exercise 12 min ✓ How Models Learn exercise 12 min ✓ Watch: Gradient Descent watch 6 min ✓ Watch: Backprop & the Chain Rule watch 5 min ✓ Backprop & Gradient Checking exercise 14 min ✓ Build an Autograd Engine exercise 16 min ✓ Train a Model with Your Engine exercise 15 min ✓ Self-Attention from Scratch exercise 16 min ✓ Watch: Attention Weights watch 5 min ✓ Watch: How Text Becomes Numbers watch 4 min ✓ What Is an LLM? exercise 11 min ✓ On-Device AI (No Cloud) exercise 10 min ✓ Prompt Engineering exercise 12 min ✓ AI Capstone: your code drives an on-device LLM exercise 20 min ✓ Concept Check: Neural-Net Intuition, LLMs & AI Capstone exercise 5 min
Module 12

Calling a Real LLM

Drive a real LLM through its actual API: build the request, read the response and tool calls, fold a token stream, and handle errors the way production code does.

✓ Build the Messages Payload exercise 12 min ✓ Read the Response exercise 12 min ✓ The Tool-Call Round Trip exercise 13 min ✓ Parallel Tool Calls & is_error exercise 13 min ✓ Fold a Token Stream exercise 12 min ✓ Handle API Errors exercise 12 min ✓ Concept Check: Calling a Real LLM exercise 5 min
Module 13

Prompt Engineering for AI Engineers

Treat the prompt as a program you build, validate, and harden in pure Python, so an LLM becomes a reliable component you control.

✓ Roles & System Prompts exercise 11 min ✓ Zero-Shot vs Few-Shot: When to Show Examples exercise 10 min ✓ Chain-of-Thought & Parsing the Final Answer exercise 11 min ✓ Structured JSON Output + Validation exercise 12 min ✓ Reusable Prompt Templates with Variables exercise 11 min ✓ Task Decomposition & Routing exercise 11 min ✓ Self-Consistency: Sample, Parse, Vote exercise 12 min ✓ Evaluating Prompts (Accuracy on a Gold Set) exercise 12 min ✓ Prompt Injection & Defenses exercise 12 min ✓ Context & Token Budgeting exercise 11 min ✓ Delimiters, Grounding & Anti-Hallucination exercise 12 min ✓ Capstone: A Production Prompt Pipeline (Ticket Triage) exercise 19 min ✓ Concept Check: Prompt Engineering exercise 6 min
Module 14

Structured Outputs & Function/Tool Calling

Make an LLM emit machine-readable, schema-valid JSON and route its tool calls to real Python functions -> the integration layer between a model and your systems.

✓ JSON Mode: Parse + Validate Against a Schema exercise 12 min ✓ Repairing Malformed Model JSON exercise 13 min ✓ Constrained / Schema-Enforced Outputs exercise 14 min ✓ Tools over MCP (Model Context Protocol) exercise 13 min ✓ Function -> Tool Schema (Introspection) exercise 12 min ✓ Tool Dispatcher / Router exercise 12 min ✓ Concept Check: Structured Outputs & Function/Tool Calling exercise 5 min
Module 15

Embeddings & Semantic Search from Scratch

Implement the vector math under every RAG and routing system -> cosine, BM25, hybrid fusion, reranking -> by hand in numpy/sklearn.

✓ Bag-of-Words Vectors exercise 10 min ✓ Cosine Similarity in numpy exercise 11 min ✓ Embeddings Intuition (Deterministic Mock Embedder) exercise 12 min ✓ Watch: Words Become Vectors watch 5 min ✓ TF-IDF Retriever Over a Doc Set exercise 12 min ✓ BM25: Sparse Lexical Ranking exercise 13 min ✓ Hybrid Search: Reciprocal Rank Fusion + Reranking exercise 13 min ✓ Watch: K-means Clustering watch 5 min ✓ Clustering for Routing & Dedup (KMeans) exercise 11 min ✓ Concept Check: Embeddings & Semantic Search from Scratch exercise 5 min
Module 16

Build a RAG Pipeline

Assemble chunking, a vector store, top-k retrieval, grounded-prompt construction, and out-of-scope refusal into a working retrieval-augmented generation engine.

✓ Watch: How RAG Works watch 5 min ✓ Chunking Long Documents exercise 12 min ✓ Embed the Chunks exercise 13 min ✓ A Tiny Vector Store exercise 13 min ✓ Out-of-Scope Refusal Guardrail exercise 11 min ✓ Assembling the Grounded Prompt exercise 12 min ✓ Concept Check: RAG exercise 6 min
Module 17

Evaluating RAG & Retrieval

Measure retrieval quality separately from generation by implementing Precision@k, Recall@k, MRR, and NDCG@k from scratch.

✓ Precision@k and Recall@k exercise 12 min ✓ Mean Reciprocal Rank exercise 11 min ✓ NDCG@k: Graded Relevance exercise 13 min ✓ Capstone: A Retrieval Eval Report exercise 16 min ✓ Concept Check: Evaluating RAG & Retrieval exercise 5 min
Module 18

LLM Evals & Testing

Build the eval harness that turns 'the prompt seems better' into numbers: assertion tests, exact-match/F1/ROUGE scorers, an LLM-as-judge parser, and a regression gate.

✓ Assertion-Based Evals (exact-match & contains) exercise 12 min ✓ Token-F1 & ROUGE-style Overlap exercise 13 min ✓ Accuracy vs Precision/Recall/F1 for Classification exercise 12 min ✓ LLM-as-Judge (Rubric Build + Verdict Parse) exercise 13 min ✓ A Prompt Regression Gate exercise 13 min ✓ Concept Check: LLM Evals & Testing exercise 5 min
Module 19

Building Agents (Tool-Use & ReAct)

Implement the reason-act-observe loop, a tool router, multi-step state, and loop guards as pure functions - the control flow under every 'AI agent', with the model mocked by a deterministic policy.

✓ The ReAct Loop Over Mock Tools exercise 13 min ✓ Intent -> Tool + Argument Extraction exercise 12 min ✓ Loop Guards: max_steps & Repeated-Action Halt exercise 12 min ✓ Agent Memory & State Accumulation exercise 14 min ✓ Planning, Reflection & Replanning exercise 16 min ✓ Concept Check: Building Agents (Tool-Use & ReAct) exercise 5 min
Module 20

Productionizing LLMs: Cost, Caching & Guardrails

Build the production layer around a model -> token/cost accounting, context budgeting, semantic and hash caching, PII redaction, and output guards -> in pure Python/pandas.

✓ Token & Cost Accounting exercise 11 min ✓ Context-Window Budgeter exercise 11 min ✓ Semantic & Hash Caching exercise 13 min ✓ PII Redaction Guardrail exercise 12 min ✓ Output Guards & Trace Aggregation exercise 13 min ✓ Prompt-Injection Defense Pipeline exercise 12 min ✓ Concept Check: Productionizing LLMs: Cost, Caching & Guardrails exercise 5 min
Module 21

Production Patterns: Retries, Async, Streaming & Memory

Build the real-world engineering every LLM app needs around the model: retry with backoff, concurrent batch calls, streaming accumulation, and multi-turn conversation memory.

✓ Retry With Exponential Backoff exercise 14 min ✓ Concurrent Batch Calls (bounded fan-out) exercise 14 min ✓ Streaming Token Accumulation exercise 13 min ✓ Multi-Turn Conversation Memory exercise 14 min ✓ Concept Check: Production Patterns: Retries, Async, Streaming & Memory exercise 5 min
Module 22

Responsible AI: Safety, Moderation & Red-Teaming

Build the safety layer a production LLM app needs: detect jailbreaks, moderate content, red-team with an adversarial suite, and calibrate refusals.

✓ Jailbreak & Prompt-Injection Detection exercise 13 min ✓ A Toxicity Moderator With an Abstain Band exercise 15 min ✓ An Adversarial Red-Team Suite exercise 15 min ✓ Calibrating Refusals: Over- vs Under-Refusal exercise 16 min ✓ Concept Check: Responsible AI: Safety, Moderation & Red-Teaming exercise 5 min
Module 23

The ML Around the LLM

Build the cheap, deterministic ML that wraps a production LLM app -> intent classification, embedding routing, and confidence thresholding -> with sklearn.

✓ Intent Classification (TF-IDF + LogisticRegression) exercise 13 min ✓ Embedding-to-Prototype Routing exercise 12 min ✓ Confidence Thresholding & Abstain exercise 11 min ✓ Cost-Aware Cascade exercise 12 min ✓ Concept Check: The ML Around the LLM exercise 5 min
Module 24

Fine-Tuning, Conceptually

Understand LoRA/DPO/PEFT conceptually and build the part you CAN run offline: instruction-tuning dataset curation, validation, dedup, and a clean train/val split.

✓ Curating an Instruction-Tuning Dataset exercise 13 min ✓ Building DPO Preference Pairs exercise 12 min ✓ Deterministic Train/Val Split exercise 11 min ✓ Memory Footprint: 4-bit vs 16-bit exercise 11 min ✓ Concept Check: Fine-Tuning, Conceptually exercise 5 min
Module 25

Projects: Build Real Things

Architect and run real programs from a near-blank file: a CLI, a log analyzer, a parser, an interpreter, and a data investigation.

✓ Project: a Task CLI exercise 18 min ✓ Project: Server Log Analyzer exercise 20 min ✓ Project: Markdown -> HTML exercise 22 min ✓ Project: a Tiny Interpreter exercise 25 min ✓ Project: Investigate a Dataset exercise 22 min ✓ Design It Yourself: A Daily Report exercise 18 min ✓ Concept Check: Projects: Build Real Things exercise 5 min
Module 26

Project: Customer Support Copilot

Build and evaluate a retrieval-augmented support assistant end to end -> the resume line "built and evaluated a RAG pipeline and eval harness".

✓ Index the Knowledge Base exercise 16 min ✓ Retrieve + Build the Grounded Prompt exercise 15 min ✓ Out-of-Scope Refusal, Citations & PII exercise 16 min ✓ Evaluate the Assistant (P@k + Answer Accuracy + Gate) exercise 18 min ✓ Lab: Watch a Real Model Ground Its Answer 7 min ✓ Concept Check: Project: Customer Support Copilot exercise 5 min
Module 27

Project: Document Intelligence Service

Turn messy documents into validated, reconciled, machine-readable data -> "built a schema-validated extraction pipeline with reconciliation + evals".

✓ Extract Fields from Messy Text exercise 16 min ✓ Validate Against a Schema + Repair JSON exercise 15 min ✓ Reconcile with Business Rules exercise 15 min ✓ Batch Eval Over Labeled Docs exercise 17 min ✓ Lab: Watch a Real Model Extract to JSON 7 min ✓ Concept Check: Project: Document Intelligence Service exercise 5 min
Module 28

Project: Production AI Gateway

Build the production layer that classifies, routes, caches, costs, and guards every request -> "built an AI gateway: classify, route, cache, guardrail, cost-account".

✓ Intent Classifier with Abstain exercise 15 min ✓ Route to a Handler + Fallback exercise 14 min ✓ Token/Cost Accounting + Caching exercise 16 min ✓ Guardrail Pipeline + Trace exercise 15 min ✓ Concept Check: Project: Production AI Gateway exercise 5 min
Module 29

Project: Tool-Using Research Agent

Implement the reason-act-observe loop, tool routing, planning and loop guards behind every "AI agent" -> "built a ReAct agent loop with tool routing, loop guards and trace eval".

✓ Tool Registry + Dispatcher exercise 15 min ✓ Wire the Agent's ReAct Loop exercise 16 min ✓ Planning + Loop Guards exercise 15 min ✓ Evaluate the Agent Trace exercise 15 min ✓ Lab: Watch a Real Model Pick a Tool 7 min ✓ Concept Check: Project: Tool-Using Research Agent exercise 5 min
Module 30

Project: Prompt Evaluation CI

Turn "the prompt seems better" into a CI gate that blocks regressions -> "built a prompt-eval harness + regression gate".

✓ Gold Set + Scorers exercise 15 min ✓ LLM-as-Judge: Rubric + Verdict Parse exercise 15 min ✓ The Regression Gate exercise 16 min ✓ Concept Check: Project: Prompt Evaluation CI exercise 5 min
Module 31

Capstone: Ship It (Package Your Project)

Turn the capstone code you wrote into a real, runnable, testable repo a hiring manager can open in 30 seconds -> the resume-ready artifact, not a notebook nobody can run.

✓ Lay Out the Repo exercise 14 min ✓ Write a Test Harness exercise 15 min ✓ README + One-Command Run exercise 15 min ✓ Concept Check: Capstone: Ship It (Package Your Project) exercise 5 min ✓ Run It For Real: From the Sandbox to a Live Model 10 min

What learners say

Leave a review

Reviews are public. Floati stores only your name, your review, and a hashed IP (to block spam) -> no account, no email, no tracking.