Generative AI Engineer Roadmap

Build reliable products on top of language models: prompts, retrieval, agents and evaluation.

7 milestones · ~185 hours total

1. 1

   LLM fundamentals

   ~15h

   Tokens, context windows, sampling, embeddings and why models fail in predictable ways.

   Practice this →
2. 2

   Prompting and structured output

   ~20h

   Clear instructions, examples, JSON schemas, validation and retries.

   Practice this →
3. 3

   Retrieval-augmented generation

   ~30h

   Chunking, embeddings, vector search, reranking and citing sources in answers.

   Practice this →
4. 4

   Tools and agents

   ~30h

   Function calling, multi-step workflows, guardrails, permissions and when not to use an agent.

   Practice this →
5. 5

   Evaluation

   ~25h

   Build test sets, automatic and human grading, regression tests for prompts and cost and latency tracking.

   Practice this →
6. 6

   Safety and privacy

   ~15h

   Prompt injection, data leakage, personal data handling and responsible release practices.

   Practice this →
7. 7

   Ship an AI product

   ~50h

   Deploy one real application with logging, rate limits, fallbacks and a feedback loop.

   Practice this →