Learn AI Engineering: The Complete Guide to Mastering AI and LLMs from Scratch Using Free Resources
The AI field moves faster than any other discipline in software engineering. New models, frameworks, and paradigms emerge weekly. For engineers trying to break in — or stay current — the challenge isn't finding resources. It's knowing which resources matter, in what order, and how they connect.
Learn AI Engineering is a comprehensive, structured roadmap created by Ashish Pratap Singh that curates the best free resources for learning AI, machine learning, deep learning, LLMs, agents, RAG, MCP, and MLOps — all organized in a logical learning path from absolute beginner to production-ready AI engineer.
With 4,975+ GitHub stars and 1,243 forks, it's one of the fastest-growing AI learning repositories, covering 15+ topics with resources from the most authoritative sources: Google, Microsoft, Anthropic, OpenAI, DeepLearning.ai, Coursera, Stanford, Fast.ai, and 3Blue1Brown.
Why This Repository Stands Out
Most "awesome lists" dump hundreds of links with no structure. Learn AI Engineering is different:
- Logical progression: Each section builds on the previous one — math foundations before ML, ML before deep learning, deep learning before LLMs, LLMs before agents
- Quality over quantity: Only the best free resources are listed — no filler, no outdated content
- 2026-current: Includes the latest paradigms: agentic coding tools (Claude Code, Codex), Model Context Protocol (MCP), and reasoning LLMs
- Practical + theoretical: Balances conceptual understanding (papers, courses) with hands-on tools (LangChain, Ollama, Cursor)
Key Stats
| Metric | Value |
|---|---|
| GitHub Stars | 4,975+ |
| Forks | 1,243 |
| License | GPL-3.0 |
| Author | Ashish Pratap Singh (ashishps1) |
| Created | April 2025 |
| Topics | agentic-ai, agents, ai, deep-learning, generative-ai, llm, machine-learning, mcp, ml, prompt-engineering, rag |
The Complete Learning Path
1. Mathematical Foundations
Every AI journey starts with math. The roadmap recommends:
- Essence of Linear Algebra — 3Blue1Brown's legendary visual series
- Probability & Statistics — Khan Academy
- Statistics Fundamentals — Josh Starmer (StatQuest)
- Mathematics for Machine Learning — Coursera (Andrew Ng)
2. Python
- AI Python for Beginners — DeepLearning.ai
3. AI & ML Fundamentals
- Machine Learning Crash Course — Google
- AI for Beginners — Microsoft
- Elements of AI — University of Helsinki
- Machine Learning Specialization — Coursera
4. Deep Learning
- Deep Learning Specialization — Coursera (Andrew Ng)
- Practical Deep Learning for Coders — Fast.ai
- Mathematics for Deep Learning — Dive into Deep Learning
Specializations: Computer Vision, Natural Language Processing (NLP), Reinforcement Learning
5. Generative AI
- The Building Blocks of Generative AI
- Generative AI for Beginners — Microsoft (21 lessons)
- Generative AI for Everyone — Coursera
6. Large Language Models (LLMs)
The largest section, covering:
- Theory: Illustrated Transformer, Sebastian Raschka's Understanding LLMs, Reasoning LLMs, Multimodal LLMs, Mixture of Experts
- Practice: Building GPT from Scratch (Andrej Karpathy), LLM Course (Hugging Face, mlabonne), Finetuning LLMs
LLM Ecosystem
| Category | Resources |
|---|---|
| Chatbots | ChatGPT, Gemini, Claude, Perplexity |
| Open Source | Llama, DeepSeek |
| APIs | OpenAI, Anthropic, Gemini, Groq |
| Frameworks | LangChain, LlamaIndex, Ollama, Instructor, Outlines |
| IDEs | Cursor, Windsurf, GitHub Copilot |
| Agentic Coding | Claude Code, Codex |
7. Prompt Engineering
- Google Prompting Essentials — Coursera
- ChatGPT Prompt Engineering for Developers — DeepLearning.ai
- Advanced Prompting Techniques — Instructor
- Getting Structured LLM Output — DeepLearning.ai
8. Retrieval-Augmented Generation (RAG)
- Introduction to RAG — Coursera
- RAG Techniques — GitHub
9. AI Agents
- A Visual Guide to LLM Agents — Maarten Grootendorst
- Agents — Chip Huyen
- AI Agents Course — Hugging Face
- Building AI Browser Agents — DeepLearning.ai
10. Model Context Protocol (MCP)
- MCP Introduction — Anthropic
- Building AI Apps using MCP — DeepLearning.ai
- MCP Course — Hugging Face
11. MLOps & Deployment
- ML in Production — Coursera
- Full Stack Deep Learning
- ML System Design — Stanford
Recommended Books
The roadmap includes 16 essential books:
| Book | Focus |
|---|---|
| Hands-On Machine Learning | Practical ML with Scikit-Learn, Keras, TensorFlow |
| Deep Learning (Goodfellow) | Theoretical deep learning foundations |
| AI Engineering | Production AI systems |
| Build a LLM from Scratch | Hands-on transformer construction |
| Prompt Engineering for LLMs | Advanced prompting techniques |
| AI Agents in Action (2nd ed.) | Agent architecture and deployment |
| Build a Multi-Agent System | Multi-agent architectures from scratch |
| Build a Reasoning Model | Reasoning model construction |
| LLMs in Production | Production deployment patterns |
| Designing ML Systems | ML system design principles |
Must-Read AI Papers
The roadmap highlights 6 foundational papers:
- Attention Is All You Need (2017) — The transformer architecture
- Generative Adversarial Networks (2014) — GANs
- GPT (2018) — Generative pre-training
- GPT-3 (2020) — Few-shot learning
- BERT (2018) — Bidirectional transformers
- Chain-of-Thought Prompting (2022) — Reasoning in LLMs
Learn AI Engineering vs Alternative Roadmaps
| Feature | Learn AI Engineering | ml-roadmap | krishnaik06 | microsoft gen-ai | awesome-gen-ai-guide |
|---|---|---|---|---|---|
| Stars | 4.9K | 2K+ | 5K+ | 75K+ | 10K+ |
| Scope | Full stack (Math→MLOps) | ML focused | 3 career paths | Gen AI only | Gen AI + research |
| Format | Curated links | Roadmap + links | Video roadmap | 21-lesson course | Papers + courses |
| LLMs | ✅ Deep | Basic | ✅ | ✅ | ✅ |
| Agents | ✅ | ❌ | ✅ | ✅ | ✅ |
| MCP | ✅ | ❌ | ❌ | ❌ | ❌ |
| Agentic Coding | ✅ Claude Code, Codex | ❌ | ❌ | ❌ | ❌ |
| Books | ✅ 16 books | Limited | Some | ❌ | ❌ |
| Papers | ✅ 6 foundational | ❌ | ❌ | ❌ | ✅ Monthly |
| MLOps | ✅ | ❌ | ✅ | ❌ | ❌ |
| Free only | ✅ | Mixed | Mixed | ✅ | ✅ |
When to Choose Each
- Learn AI Engineering: Best all-in-one roadmap covering the complete journey from math through production deployment, with the latest 2026 topics (MCP, agentic coding). Ideal for self-directed learners who want a structured path.
- ml-roadmap: Streamlined ML-only path with experimental AI-assisted features. Good for focused ML learning.
- krishnaik06: Three distinct career paths (Data Scientist, Gen AI Engineer, Agentic AI Developer). Best for those wanting career-specific guidance.
- microsoft generative-ai-for-beginners: Hands-on 21-lesson course. Best for structured, project-based learning focused specifically on generative AI.
- awesome-generative-ai-guide: Research-oriented with monthly paper curation. Best for staying current with the latest academic developments.
Use Cases
🎓 Career Transition to AI Engineering
Follow the roadmap from math foundations through MLOps for a complete transition path with no paid resources required.
📚 Self-Paced Study
Cherry-pick sections based on your current knowledge — skip math if you have the background, jump to LLMs if you understand deep learning.
🏢 Team Onboarding
Use as a standard onboarding curriculum for new AI engineering hires — the structured progression ensures everyone builds the same foundational knowledge.
🔬 Research Preparation
The must-read papers section and book recommendations provide the theoretical depth needed for research-oriented work.
🛠️ Staying Current
The LLM ecosystem section (chatbots, IDEs, agentic tools, MCP) provides a quick overview of the current production landscape.
Frequently Asked Questions
Is this for beginners or advanced engineers?
Both. The roadmap starts from mathematical foundations and Python, but also covers advanced topics like reasoning LLMs, MCP, and multi-agent systems.
Are all resources free?
Yes. Every course, video, and tutorial linked is freely accessible. The books section is the only area with paid resources.
How long does the complete path take?
Depends on your pace and background. With dedicated study, the core path (math through LLMs) can be completed in 3-6 months. Adding agents, MCP, and MLOps extends to 6-12 months.
Is this kept up to date?
Yes. The repository includes 2026-current tools like Claude Code, Codex, and MCP courses from Hugging Face and DeepLearning.ai.
Conclusion
Learn AI Engineering fills a critical gap in the AI education landscape. While most resources are either too narrow (single-topic courses) or too broad (massive awesome lists), this roadmap provides a structured, progressive path through the entire AI engineering stack — from linear algebra to production deployment.
With 4,975+ stars and resources from the most authoritative sources in AI (Google, Microsoft, Anthropic, Stanford, DeepLearning.ai), it's the single best starting point for anyone serious about AI engineering in 2026.
The most valuable aspect? It's opinionated about quality. Rather than listing everything, it lists only the best — saving you countless hours of evaluation and comparison.