Build Fast with AI logo
RoadmapProjectsCareer PathsCheckpoints
Start Level 0
  1. Home
  2. GenAI Roadmap
Build Fast with AI logoFree · Self-paced · Project-based

Generative AI Roadmap 2026: From LLM Basics to AI Agents

This Gen AI roadmap takes you from your first LLM API call to production RAG systems and AI agents in 7 levels. Levels 0 to 5 take about 129 hours, roughly 16 weeks at 8 hours a week, and each one ends with checkpoint tasks you can verify yourself.

By the Build Fast with AI team·Published April 10, 2026·Updated November 19, 2026

Start Learning Explore Career Paths
Progressive learning
7 Levels
For Levels 0–5
~129 Hours
Hands-on checkpoints
35 Tasks
Run free in Colab
13 Notebooks
🎯

Clear progression

7 levels in a fixed order, from Python setup to fine-tuning, so you always know what to learn next and why.

💻

Hands-on at every level

You build with the OpenAI, Anthropic and Gemini APIs, LangChain, Pinecone and Qdrant, FastAPI, Docker and Promptfoo. 13 of the 47 projects open in Google Colab.

📈

Four career paths

After Level 5, specialize as an AI App Developer, AI Engineer, ML Researcher or AI Product Manager, with four portfolio projects each.

Jump toThe roadmap at a glanceLevel by levelHow long it takesWhy this orderNotebooks you can run todayCareer pathsFAQKeep learning

The Gen AI roadmap at a glance

Seven levels, each building on the one before. Times assume about 8 hours of study a week.

LevelWhat you learnTimeDifficultyKey tools
Level 0Environment & Foundations4 hours (~4h)BeginnerPython, venv or uv, Jupyter
Level 1LLM Fundamentals & APIs1-2 weeks (~15h)BeginnerOpenAI API, Anthropic Claude API, Gemini API
Level 2Building Simple GenAI Applications2-3 weeks (~20h)IntermediateLangChain, Pydantic, Streamlit
Level 3RAG Systems & Intelligent Agents3-4 weeks (~25h)IntermediateEmbeddings, Pinecone, Weaviate
Level 4Production Systems & Deployment3-4 weeks (~30h)AdvancedFastAPI, Docker, PostgreSQL
Level 5Advanced GenAI Techniques4-5 weeks (~35h)AdvancedLoRA and QLoRA, Hugging Face, Gemma
Level 6Choose Your Specialization Path8-14 weeks (~60h)AdvancedAgno, OpenAI Agents SDK, AgentOps

Level by level

Start at Level 0 if you have never called an LLM API from Python. If you have, jump to Level 1 or to whichever level covers what you can't do yet.

0Beginner

Environment & Foundations

Set up Python, Jupyter, Git and your first LLM API key, then run quickstarts for OpenAI Agents, GPT-5.4 mini/nano and Gemini Flash Live.

4 hours · ~4h
1Beginner

LLM Fundamentals & APIs

Understand how LLMs work and start building with GPT-5.5, Claude Opus 4.7, and Promptfoo-based eval workflows.

1-2 weeks · ~15h
2Intermediate

Building Simple GenAI Applications

Create interactive GenAI apps using GPT-5.4, Gemini 3.1 Flash-Lite, and Gemma 4 for structured outputs and routing.

2-3 weeks · ~20h
3Intermediate

RAG Systems & Intelligent Agents

Build retrieval-augmented generation systems and multi-step autonomous agents with modern tracing and handoffs.

3-4 weeks · ~25h
4Advanced

Production Systems & Deployment

Deploy GenAI applications at scale with monitoring, evaluation, benchmarking, and guardrails.

3-4 weeks · ~30h
5Advanced

Advanced GenAI Techniques

Master long-context reasoning, multimodal systems, and advanced optimization with GPT-5.5, Claude Opus 4.7, and Gemma 4.

4-5 weeks · ~35h
6Advanced

Choose Your Specialization Path

Deep dive into app development, engineering, research, or product using the latest agents, benchmarks, and coding models.

8-14 weeks · ~60h

How long does it take to learn generative AI?

About 4 months part-time. Levels 0 to 5 add up to roughly 129 hours: 4 hours of setup, then 15, 20, 25, 30, 35 hours for Levels 1 to 5. At 8 hours a week that is about 16 weeks; at 15 hours a week it is closer to 9. A specialization adds 8 to 14 weeks on top.

The fastest route through isn't skipping levels. It's finishing the 35 checkpoint tasks, because that's where you find out what you only thought you understood.

Levels 0–5
~129h
At 8 h/week
16 weeks
At 15 h/week
9 weeks

Why the roadmap is in this order

Our take: most GenAI projects that fail don't fail because of the model. They fail on retrieval that returns the wrong chunk, prompts nobody tested, or token bills nobody tracked. So the roadmap puts testing and debugging early and fine-tuning late. Want to build an agent in week one? You can, but you'll spend week two figuring out why it loops.

  • 1

    Call the raw API before any framework.

    LangChain arrives in Level 2, after Level 1 has shown you what a request, a token count and a bill look like.

  • 2

    Learn RAG before you lean on agents.

    An agent is a loop of LLM calls. If you can't debug one retrieval step, you won't debug ten.

  • 3

    Test prompts like code from day one.

    Promptfoo shows up in Level 1 and again in Level 4 for exactly this reason.

  • 4

    Leave fine-tuning for Level 5.

    Better prompts plus retrieval fix most of the problems people reach for fine-tuning to solve, and they're easier to change later.

Notebooks you can run today

13 runnable notebooks, each tied to the level where it fits. Every one opens in Google Colab, so you don't need to install anything to try it.

GPT-5.4 Mini/Nano Cookbook

Notebook

Learn the latest OpenAI mini and nano model setup, structured outputs, and fast prompt patterns.

BeginnerLevel 0–1#openai#gpt-5.4-mini

OpenAI Agents Quickstart

Notebook

Build triage, handoff, and tool-calling workflows with the OpenAI Agents SDK.

BeginnerLevel 0–3#agents#handoffs

Gemini 3.1 Flash Live Preview

Notebook

Explore low-latency multimodal and voice-first workflows with Gemini 3.1 Flash Live Preview.

BeginnerLevel 0–2#gemini#audio

GPT-5.5 Cookbook

Notebook

Work with GPT-5.5 and GPT-5.5 Pro for long-context reasoning, coding, and agentic workflows.

IntermediateLevel 1–5#openai#gpt-5.5

Claude Opus 4.7 Cookbook

Notebook

Use adaptive thinking, effort controls, and long-horizon coding patterns with Claude Opus 4.7.

IntermediateLevel 1–5#anthropic#claude-opus-4.7

Promptfoo GPT-5.4 Evaluation Cookbook

Notebook

Compare frontier models with cost, latency, and red-team checks using Promptfoo.

IntermediateLevel 1–4#promptfoo#evals
Show 7 more notebooks

Gemma 4 Cookbook

Notebook

Build multimodal, multilingual, and hybrid-thinking applications with Gemma 4.

IntermediateLevel 2–5#gemma#multimodal

Gemini 3.1 Flash-Lite Guide

Notebook

Route simple tasks to Flash-Lite and harder ones to Pro with a practical hybrid router.

IntermediateLevel 2–3#gemini#routing

GPT-5.4 Cookbook

Notebook

Ship structured outputs, agent tools, and long-context workflows with GPT-5.4.

IntermediateLevel 2–4#openai#gpt-5.4

Opik LLM Evaluation and Monitoring

Notebook

Track traces, metrics, and evaluation runs for LLM and agent workflows.

IntermediateLevel 3–4#opik#monitoring

AgentOps Evaluation

Notebook

Add replay analytics, benchmarking, and monitoring to multi-agent systems.

IntermediateLevel 3–6#agentops#monitoring

LLM Benchmark Comparison

Notebook

Compare frontier models across reasoning, coding, and speed-focused tasks.

AdvancedLevel 4–6#benchmarking#comparison

Qwen3 Coder Next with Agno

Notebook

Build a coding-focused agent with a modern OpenRouter model and the Agno framework.

AdvancedLevel 6#qwen#agno
See all 47 projects and briefs

Choose a career path after Level 5

Levels 0 to 5 are shared. Level 6 splits into four tracks of 8 to 14 weeks, each ending with four portfolio projects.

🚀

AI App Developer Roadmap

Specialize in building consumer-facing GenAI applications. Master frontend frameworks, real-time communication, and product-focused development.

8-10 weeks · 4 projects · Senior AI/Full-Stack Engineer

React/Vue for modern frontendsReal-time communication (WebSockets)UI/UX for AI applications
Read the roadmap
⚙️

AI Engineer Roadmap

Specialize in building scalable, efficient AI systems. Focus on system design, optimization, and infrastructure.

8-12 weeks · 4 projects · ML/AI Systems Architect

System architecture and designPerformance optimizationDistributed systems basics
Read the roadmap
🔬

ML Researcher Roadmap

Run original experiments, publish papers and reproduce the newest methods in retrieval, agents and fine-tuning.

10-14 weeks · 4 projects · ML Research Scientist

Research methodologyAcademic writing and papersExperimentation frameworks
Read the roadmap
📊

AI Product Manager Roadmap

Lead AI products and strategy. Master customer development, metrics, and go-to-market strategy.

8-10 weeks · 4 projects · Product Manager / AI Product Lead

Customer developmentProduct metrics and analyticsMarket positioning
Read the roadmap

FAQ

Generative AI roadmap FAQ

What is a generative AI roadmap?

A generative AI roadmap is an ordered plan for learning to build with models such as GPT, Claude and Gemini. This one has 7 levels: setup, LLM basics, simple apps, RAG and agents, production, advanced techniques, and a career specialization.

How long does it take to learn generative AI?

Levels 0 to 5 take about 129 hours. At 8 hours a week that is roughly 16 weeks, or about 4 months; at 15 hours a week it is closer to 9 weeks. A specialization adds another 8 to 14 weeks.

Can a complete beginner follow this Gen AI roadmap?

Yes, if you know basic Python: variables, functions, lists and installing packages. Level 0 sets up your environment in about 4 hours, and you do not need machine learning math until the fine-tuning work in Level 5.

Is the GenAI roadmap free?

Yes. Every level, checkpoint task and notebook is free, and the 13 notebooks run in Google Colab. You only pay for the LLM API calls you make, which is usually a few dollars in the early levels.

What should I learn first: prompt engineering, RAG or AI agents?

Prompt engineering and LLM APIs first (Level 1), then simple apps (Level 2), then RAG and agents together (Level 3). Agents are loops of LLM calls, so they are much easier to debug once you can test a single prompt and a single retrieval step.

Do I need machine learning or math to learn generative AI?

Not for Levels 0 to 4, which use models through APIs. Level 5 introduces fine-tuning with LoRA, where some linear algebra helps, and the ML Researcher track goes deeper into the math.

Which careers does this roadmap prepare you for?

After Level 5 you pick one of four tracks: AI App Developer, AI Engineer, ML Researcher or AI Product Manager. Each takes 8 to 14 weeks and ends with four portfolio projects for that role.

Read next

Keep learning

  • What is generative AI?
  • What is an LLM?
  • What is RAG?
  • What is an AI agent?
  • Generative AI interview questions
  • AI project ideas with source code
Build Fast with AI logoBuild Fast with AI

A free, project-based roadmap for learning generative AI, from your first API call to production systems.

Levels

  • Level 0: Environment & Foundations
  • Level 1: LLM Fundamentals & APIs
  • Level 2: Building Simple GenAI Applications
  • Level 3: RAG Systems & Intelligent Agents
  • Level 4: Production Systems & Deployment
  • Level 5: Advanced GenAI Techniques
  • Level 6: Choose Your Specialization Path

Career paths

  • AI App Developer Roadmap
  • AI Engineer Roadmap
  • ML Researcher Roadmap
  • AI Product Manager Roadmap
  • All projects and notebooks
  • Checkpoint tasks

Keep learning

  • AI concepts explained
  • AI glossary
  • AI project ideas
  • AI interview prep
  • Blog
  • Notebooks on GitHub

© 2026 Build Fast with AI. All rights reserved.

Privacy PolicyTerms