The Neural Mastery Roadmap

One path through everything on this site, from zero to job-ready. Each section links to its own detailed roadmap — this page is the map of the map.

Have a specific goal already (ML Engineer, LLM/GenAI Engineer, AI Research Engineer, AI Systems/Infra Engineer)? See Suggested Paths for a shorter, goal-specific route through the same sections instead.

Beginner Track — Foundations

1. CS Fundamentals for AI Engineers — operating systems, concurrency, networking, distributed systems, Linux/Git, software engineering practice
2. Python Engineering for AI — OOP & functional Python, decorators & generators, async/concurrency, memory management, packaging, testing, tooling
3. Mathematics for AI — linear algebra, calculus, probability, algorithms & data structures
4. Machine Learning — classical algorithms, evaluation, the ML workflow

Intermediate Track — Deep Learning & Modern AI

5. Deep Learning — neural networks, CNNs, sequence models, attention & Transformers
6. Computer Vision — image fundamentals, the vision task taxonomy, modern/multimodal vision
7. NLP — classical NLP, word embeddings, the task taxonomy underneath modern LLMs
8. Speech & Audio AI — signal processing, spectrograms/MFCCs, ASR, TTS, speaker recognition, diarization
9. LLMs & GenAI — foundation models, training pipelines, prompt engineering, RAG, evaluation
10. Databases — relational, vector (ChromaDB & friends), graph (Neo4j) — needed in parallel with RAG

Advanced Track — Systems & Agents

11. Agents — agent architectures, MCP, A2A, multi-agent systems
12. AI Evaluation — benchmark design, golden datasets, LLM-as-judge, agent evaluation, human & adversarial evaluation
13. ML System Design — the 9-step framework, recommendation/search/GenAI system case studies
14. MLOps — data/experiment versioning, pipelines, containers, Kubernetes, CI/CD, monitoring, LLMOps & RAGOps
15. Frameworks — PyTorch, LangChain, MCP SDKs, serving & LLMOps tooling

Specialization Track — Safety, Theory & Domains

16. AI Security — the OWASP LLM Top 10, prompt injection, model extraction, data poisoning, supply-chain attacks
17. AI Safety & Alignment — RLHF/Constitutional AI, reward hacking, specification gaming, scalable oversight
18. Interpretability — SHAP/LIME, saliency/attention analysis, probing, sparse autoencoders, mechanistic interpretability
19. Reinforcement Learning — MDPs, Bellman equations, Q-learning, policy gradients, PPO/SAC/TD3
20. Graph ML — GNNs, graph embeddings, knowledge graphs, Graph RAG
21. AI for Science — drug discovery, protein structure prediction, genomics, materials science
22. Domain AI Applications — healthcare AI in depth, plus finance, semiconductor, cybersecurity, robotics, manufacturing

Build Track

23. Research Engineering — how to read AI papers, and the paper timeline connecting this site's architectures historically
24. Build From Scratch — implement autodiff, attention, a GPT, a vector database, RAG, an agent, yourself
25. Projects — a six-level project ladder from a tiny afternoon build to a frontier capstone

Job-Ready Track

26. Interview Prep — DSA, ML coding, knowledge Q&A, system design, behavioral

How to use this roadmap

• Each section's roadmap is a checklist — work through it top to bottom, and don't skip the "Common Problems & SOTA Solutions" parts. Interviews (and real jobs) test whether you know why a technique exists, not just its name.
• The math and ML foundations aren't optional even if you only care about LLMs/agents — attention is linear algebra, RLHF is optimization, and evaluation is statistics.
• Revisit Interview Prep continuously, not just at the end — it's the fastest way to find your weak spots early.

Want more practice?

See Open Source AI Resources for a running list of external practice sites and interview-prep resources.