A curriculum and field guide for building, evaluating and operating AI systems.
Evidence review: October 9, 2026. The latest job-data snapshot inspected is September 23, 2026. Review dates and source publication/collection dates are separate. Some long-form legacy claims still need source-level verification; see the audit.
Section review: October 17, 2026. The AI Evaluation and AI Security & Safety sections were added with all references verified on that date.
- From foundations to advanced engineering — competency stages, real-data projects, measurable exit criteria and expert review questions.
- Evidence and real datasets — a reproduced job-posting analysis, dataset limitations, dated benchmark references and reproduction commands.
- Evidence policy — distinguish measurements, source-reported claims, design targets and teaching illustrations.
- Repository audit — what was checked, corrected and left unresolved.
The reproduced source contains 8,051 stored job observations across nine scrape folders, including 1,086 records in the September 23, 2026 folder. These are counts from a selected job-board dataset, not worldwide demand estimates. Structured skills and role labels are LLM-derived annotations; the analysis and input fingerprint make the counting reproducible without certifying every annotation.
Advanced practice requires a baseline, versioned data, held-out evaluation, failure analysis, authorization checks, measured operating costs and a recovery plan. The curriculum retains its introductory explanations; the advanced path connects them to reviewable engineering artifacts.
Two complete roadmaps for becoming an AI engineer, plus a full interview-preparation track — structure adapted verbatim from the CampusX AI Engineering Roadmap 2026, with every learning resource replaced by free worldwide alternatives:
- Roadmap hub — which one to follow — comparison, and how important each topic is for interviews vs the real-world job
- Full Roadmap — 16 levels · 7 phases · 286 sessions · 14 projects; 100% coverage from software fundamentals to interview prep
- Accelerated Roadmap — 100 Days of AI Engineering — 6 modules · 100 days; the core ~80% you need to land your first AI engineering job
- The 100 GenAI / AI Engineer Interview Questions — copied exactly, with strong-answer outlines and a practice loop
- 12 portfolio projects — full specifications with "done when" criteria and a target-role selection table
- 101 — LLM Literacy — first-principles guide to transformer mechanics, inference sampling, prompt & context engineering
- 102 — AI Application Development — orchestration, vector databases, complete RAG curriculum, memory systems, and fine-tuning
- 103 — Agentic AI — autonomous planning loops, Model Context Protocol (MCP), agentic frameworks, and multi-agent orchestration
- 104 — LLMOps — deployment architectures, observability, safety guardrails, governance, and enterprise scaling
- 105 — AI Evaluation — evaluation fundamentals, LLM-as-a-Judge, agent benchmarks and the integrity crisis, RAG evaluation practice, and eval engineering in CI/CD
- 106 — AI Security & Safety — threat modeling (OWASP LLM + Agentic Top 10), prompt injection defense, red teaming, agent & MCP security, AI alignment, and responsible AI governance
- My vision of the role — how I see AI engineering, comparison with DS/ML/DE roles, CRISP-DM for AI
- Skills analysis — reproduced annotation counts, limitations and a competency rubric
- Responsibilities — engineering ownership and reviewable deliverables
- Use cases — application patterns, data requirements and measurable outcomes
- Reality vs. job postings — what candidates experience vs. what's advertised
- Worldwide job market analysis — global geographical demand, compensation profiles, and company archetypes
- Interview process — common patterns, step counts, time estimates, AI use in hiring, key takeaways
- Interview questions — practice topics and individually attributed legacy reports
- Theory — LLMs, RAG, agents, ML fundamentals, company-specific questions
- Coding — coding round formats, DSA problems, ML implementation exercises
- Project deep dive — presentation rounds, follow-up probes, what interviewers evaluate
- AI system design — system design for AI applications
- Behavioral — values, leadership, problem-solving
- Home assignments — assignment preparation and public examples; verify each source
- Skills that get you hired — baseline expectations, differentiators, and portfolio strategy
- After the interview — handling offers, rejections, and salary negotiation
- Interview trends — AI cheating, AI-proctored rounds, context engineering interviews
- Global compensation & negotiation guide — compare dated local evidence and written offer terms
- Resume & portfolio playbook — ATS optimization, high-impact bullet formulas, and 60-second GitHub repo audit checklist
- Production system design case studies — illustrative design exercises and validation requirements
- General learning path — what to learn and in what order
- From Data Engineer — smoothest transition, 3-4 months
- From Data Scientist — evaluation is your superpower, add engineering
- From ML Engineer — easiest transition, replace model call with API call
- From Backend Engineer — 2-3 months, add AI on top of engineering
- From Frontend Engineer — backend first, then AI, unique full-stack advantage
- Project ideas and built projects — project specifications with evaluation guidance, plus the five built project repositories (101–105)
Curated collection of resources compiled while researching content for this field guide:
- Practitioner interview stories
- AI system design guides
- Company engineering blogs
- Books and courses
- Case study collections
See awesome.md for the list.
Supplemental notes on AI engineering careers. Recording, speaker and survey claims in these notes have not been independently verified; see their evidence status.
- A Day of an AI Engineer — the practical reality of the role
- Defining the AI Engineer Role — role scope and engineering responsibilities
- The Interview Process — real hiring trends, technical questions, and live coding challenges
- Take-Home Assignments — analyzing real assignments and building production-ready solutions
This guide builds on the foundational work by Alexey Grigorev in the original AI Engineering Field Guide. The October 9 review links to a pinned upstream dataset and adds a reproducible counting method. Upstream annotations and third-party data retain their own limitations and rights; this repository does not claim a newly collected worldwide dataset.
Building • Learning • Researching • Sharing
⭐ If this repository helped you, consider giving it a star! ⭐
Engineering principle: Robust AI engineering requires deterministic software engineering surrounding probabilistic model outputs. Focus on evaluation suites, reproducible benchmarking, and production monitoring to ensure reliable business outcomes.