AI Engineer Roadmap

The path from "I called the OpenAI API once" to "I got the job" - and you practice each step on a real broken AI system in a live cloud workspace. Every skill links to a live ticket you fix.

Your progress0 of 12 skills

LLM API Basics

Calling an LLM

Restore a broken LLM API integration.

Practice live: Restore a Broken LLM API Integration → Parsing model output

Strip the markdown fence so JSON parses.

Practice live: Parse JSON From an LLM (Strip Markdown Fences) →

Prompt Engineering

Classification prompts

Tame a sentiment classifier with a better prompt.

Practice live: Improve an LLM Sentiment Classifier Prompt → Structured tasks

Ship a support-ticket classifier that holds up.

Practice live: Build an LLM Support Ticket Classifier →

Reliability & Cost

Resilience

Harden an LLM pipeline against API failures.

Practice live: Harden an LLM Pipeline Against API Failures → Caching responses

Cache LLM responses in Redis to cut cost.

Practice live: Cache LLM Responses in Redis →

Embeddings & Retrieval

Generating embeddings

Embed a doc set with the OpenAI API.

Practice live: Generate Text Embeddings With OpenAI → Search strategy

Design the semantic-search embedding strategy.

Practice live: Design a Semantic Search Embedding Strategy →

RAG

End-to-end RAG

Wire chunking, retrieval, and generation together.

Practice live: Build an End-to-End RAG Pipeline → Vector-powered retrieval

Power RAG search with a vector store.

Practice live: Power RAG Search With a Vector Database →

Evaluation & Routing

Eval harness

Build an LLM eval harness with golden sets.

Practice live: Build an LLM Evaluation Harness (Golden Set) → Model routing

Route requests to cheaper models with escalation.

Practice live: Route LLM Requests to Cheaper Models →

Stop reading roadmaps. Start doing them.

Every skill above is a real broken system in a live workspace. Fix it, and it lands on your portfolio.

Start free →