AI Engineer Roadmap

Build products on top of LLMs: prompting, RAG, agents and evals.

1. 1

   Foundations

   Enough programming and ML intuition to work with models.

   • Python or TypeScript
   • HTTP APIs & JSON
   • How neural nets learn
   • Linear algebra basics(optional)
   • articlePython tutorial
   • video3Blue1Brown: Neural networks
2. 2

   LLM fundamentals

   What a model sees and what it costs.

   • Tokens & context window
   • Sampling (temperature)
   • Model choice & pricing
   • Transformers(optional)
   • courseHugging Face LLM course
3. 3

   Prompting

   Get reliable output from clear instructions.

   • System prompts
   • Few-shot examples
   • Structured output (JSON)
   • articleClaude docs
4. 4

   Model APIs

   Call models from your app.

   • Messages API
   • Streaming
   • Prompt caching
   • Batch processing(optional)
   • articleClaude docs
5. 5

   RAG

   Answer from your own data instead of the model's memory.

   • Embeddings
   • Chunking
   • Vector search
   • Reranking(optional)
   • Hybrid search(optional)
6. 6

   Tools & Agents

   Let the model act: call functions, loop until done.

   • Tool use
   • Agent loop
   • MCP servers
   • Multi-agent systems(optional)
   • articleModel Context Protocol
7. 7

   Evals & Production

   Measure quality, then ship it safely.

   • Eval sets
   • LLM-as-judge
   • Guardrails & safety
   • Monitoring & cost
   • Fine-tuning(optional)