About an hour, once. Commands are for macOS and Linux; on Windows use WSL. Each step ends with a check so you know it worked.
Set up VS Code, Cursor or Claude Code ↓One fast tool for Python versions, virtual environments and packages. Every project here uses it.
$ curl -LsSf https://astral.sh/uv/install.sh | sh
$ uv init my-ai-project && cd my-ai-project
$ uv add openai python-dotenv
Check: uv run python -c "import openai; print(openai.__version__)" prints a version.
VS Code (or Cursor) with the Python extension, and one GitHub repo per project so reviewers can see your history.
$ git init && git add . && git commit -m "start"
$ gh repo create my-ai-project --public --source . --push
Check: The repo is visible on your GitHub profile.
OpenAI is the quickest start. Azure OpenAI or AWS Bedrock match what most enterprise job ads ask for.
$ echo "OPENAI_API_KEY=sk-..." >> .env
$ echo ".env" >> .gitignore
# Azure: AZURE_OPENAI_ENDPOINT, AZURE_OPENAI_API_KEY and a deployment name from the portal
# Bedrock: aws configure, then enable model access in the Bedrock console
Check: A one-line script returns a model reply, and .env is ignored by Git.
Ollama runs small open models on your laptop, so you can test without spending credits.
$ brew install ollama # or the installer from ollama.com
$ ollama pull llama3.2
$ ollama run llama3.2 "hello"
Check: The model answers in your terminal.
A real database with vector search, the same kind many teams run in production.
$ docker run -d --name pg -e POSTGRES_PASSWORD=dev -p 5432:5432 pgvector/pgvector:pg16
$ docker exec -it pg psql -U postgres -c "CREATE EXTENSION vector;"
Check: Running \dx in psql lists the vector extension.
Reviewers click before they read. Streamlit turns a script into a web page in minutes.
$ uv add streamlit
$ uv run streamlit run app.py
Check: Your app opens at localhost:8501.
Pick a tool and your computer. Each guide ends with a hello-world that proves it works.
A free editor from Microsoft. Most tutorials and job ads assume it, and Cursor is built on it.
$ code --version
$ mkdir hello-editor && cd hello-editor
$ code .
$ python3 hello.py
Check: The terminal prints: hello from VS Code
Stable 1.145 was the newest release listed on the VS Code updates page. Checked 10 Nov 2026. Source: VS Code docs: setup