Suggested Paths

The Neural Mastery Roadmap is the complete, zero-to-job-ready path through every section — the right choice if you genuinely want it all. Most people don't start there with a blank goal, though; they have a specific target role in mind. These four paths reorder and trim the same 26 sections around four common goals, so you're not working through Graph ML and AI for Science on your way to an LLM engineering job, or through Speech & Audio on your way to an ML systems role.

Each path lists sections in order, with a one-line reason for why it's there — skip a hop if you already know it, don't skip it just because it looks unfamiliar.

🧮 ML Engineer

Classical ML through production systems — the role most "AI Engineer" job postings actually mean.

1. Mathematics for AI — linear algebra and probability are the language every algorithm below is written in.
2. Machine Learning — the core toolbox: regression, trees, boosting, clustering, evaluation.
3. Deep Learning — when classical ML tops out, this is what replaces it.
4. ML System Design — the difference between "I trained a model" and "I shipped a system."
5. MLOps — versioning, pipelines, monitoring — how a model survives contact with production.
6. AI Evaluation — you can't improve what you can't measure correctly.
7. Interview Prep — DSA and ML coding rounds specifically test this exact stack.

🤖 LLM / GenAI Engineer

Building products on top of foundation models — RAG, agents, and the infrastructure around them.

1. Mathematics for AI — just enough linear algebra/probability to not treat attention as magic.
2. Deep Learning — specifically through attention and Transformers; that's the load-bearing chapter.
3. NLP — word embeddings and the task taxonomy underneath every LLM application.
4. LLMs & GenAI — foundation models, training, prompting, RAG — the center of this whole path.
5. Databases — specifically vector databases; RAG doesn't work without one.
6. Agents — where GenAI products are actually headed — tool use, MCP, multi-agent systems.
7. AI Evaluation — LLM-as-judge and agent evaluation are genuinely different from classical ML metrics.
8. AI Security — prompt injection and the OWASP LLM Top 10 are now table-stakes knowledge, not optional.
9. Interview Prep — system design rounds increasingly ask you to design a RAG or agent product specifically.

🔬 AI Research Engineer

Reading papers and reproducing results, not just calling APIs.

1. Mathematics for AI — go deep here; research code assumes you can follow a derivation, not just call .fit().
2. Machine Learning and Deep Learning — the shared vocabulary every paper is written in.
3. Research Engineering — how to actually read a paper, plus the historical lineage connecting architectures to each other.
4. Build From Scratch — implementing autodiff, attention, and a GPT yourself is the fastest way to stop taking a paper's claims on faith.
5. Interpretability — a large, active research area in its own right, and directly useful for understanding why your own experiments behave the way they do.
6. Reinforcement Learning or Graph ML — pick whichever matches your specific research interest; both are genuinely separate subfields with their own math.
7. Projects — specifically the top of the ladder: a real capstone is what a research portfolio actually needs.

🏗️ AI Systems / Infrastructure Engineer

Making AI systems fast, reliable, and cheap at scale — less modeling, more engineering.

1. CS Fundamentals for AI Engineers — operating systems, concurrency, and distributed systems fundamentals this whole path builds on.
2. Python Engineering for AI — async, memory management, and packaging matter far more here than in a notebook-only workflow.
3. Deep Learning — specifically the GPU/CUDA and distributed-training material; that's what you're actually operating.
4. ML System Design — the 9-step framework and the case studies are exactly this role's day-to-day.
5. MLOps — go deep here: Kubernetes, CI/CD, monitoring, LLMOps/RAGOps at scale is most of the job.
6. Databases — relational, vector, and graph — you'll be responsible for all three eventually.
7. Frameworks — the serving and tooling layer connecting everything above.
8. AI Security — infrastructure is the last line of defense against a lot of AI-specific attack classes.
9. Interview Prep — specifically system design practice; that's the round this role lives or dies on.

Not sure which fits? Check the Learning Path map for how these sections depend on each other regardless of which goal you pick — Foundations and (usually) Models come first no matter what.