AI Engineer Roadmap

Most self-taught learners take 9-13 months studying part-time (10-15 hours/week), or 5-7 months full-time. It's longer than the web development roadmaps because it covers math, classical ML, deep learning and modern LLM engineering — four fields that used to be separate careers.

No, but Phase 2 (Math & Statistics for AI) is not optional — you need working knowledge of linear algebra, calculus and probability to actually understand what your models are doing, not just call `.fit()` on them. You don't need a math degree; you need to be comfortable with the specific concepts this roadmap covers.

This roadmap deliberately builds classical ML (Phases 5-6) and deep learning fundamentals (Phases 7-8) before LLMs (Phase 11 onward). Skipping straight to prompting LLMs gets you shipping features fast but leaves gaps that show up the moment something needs fine-tuning, evaluation, or debugging — which is exactly what employers screen for.

Yes, completely free with no login or signup required. Your progress is saved locally in your browser so you can track your checklist across visits.

For most of it, no — classical ML, small deep learning experiments, and all the LLM API/RAG/agent phases run fine on a laptop or free-tier cloud notebook (Google Colab, Kaggle). You'll only need real GPU access for Phase 13 (Fine-Tuning), and even that's manageable with QLoRA on a single consumer GPU or a rented cloud GPU for a few hours.

The lines are blurry and job titles vary by company, but broadly: ML Engineers historically focus on training and deploying custom models (Phases 5-10 of this roadmap), while AI Engineers increasingly focus on building products with foundation models — prompting, RAG, agents and fine-tuning existing LLMs (Phases 11-17). This roadmap deliberately covers both, since most real AI engineering roles now expect familiarity with each.

Aim for 3-4 genuinely complete projects that span different parts of the stack — for example, one classical ML pipeline with real deployment, one RAG chatbot with a real knowledge base, and one AI agent that automates a real workflow. A RAG chatbot with source citations and honest 'I don't know' handling demonstrates more real skill than five notebook-only Kaggle scripts.

It covers the technical breadth companies expect, but landing a job also depends on your portfolio quality, DSA/system-design interview performance, and your ability to explain modeling trade-offs clearly — which is exactly what Phases 19-22 (System Design, DSA, Portfolio Projects, Interview Preparation) are built to prepare you for.