Working environment
- Jupyter / JupyterLab — interactive notebooks for exploration
- VS Code — editor for notebooks and production code
- Google Colab — free hosted notebooks with GPU access
- Git & GitHub — version control and collaboration
Data analysis & visualisation
- pandas & NumPy — data manipulation and numerical computing
- SQL (PostgreSQL, MySQL) — querying data where it lives
- Matplotlib, Seaborn, Plotly — charts in Python
- Power BI & Tableau — business dashboards and reporting
- Excel — still the quickest tool for many ad-hoc analyses
Machine learning
- scikit-learn — classic ML: regression, classification, clustering, pipelines
- XGBoost / LightGBM — strong performers on tabular data
- PyTorch & TensorFlow — deep learning frameworks
Generative AI & LLMs
- LLM APIs (e.g. Anthropic, OpenAI, Google) — hosted models via API
- Hugging Face — open models, datasets and the transformers library
- LangChain & LlamaIndex — building blocks for LLM and RAG apps
- Ollama — run open models locally
Agents & orchestration
- Tool / function calling — let a model call your functions and APIs
- LangGraph, CrewAI — frameworks for multi-step and multi-agent workflows
- Model Context Protocol (MCP) — a standard way to connect models to tools and data
Vector databases
- FAISS — fast similarity search library
- Chroma — lightweight embedded vector store
- pgvector — vector search inside PostgreSQL
- Pinecone, Weaviate, Qdrant — managed / scalable vector databases
Deploying & MLOps
- FastAPI / Flask — serve models as APIs
- Streamlit / Gradio — quick interactive demos
- Docker — package apps with their dependencies
- MLflow — track experiments and models
- Cloud platforms (AWS, Azure, GCP) — scalable compute and managed AI services
How to choose
Don't try to learn everything at once. Master Python, SQL and one visualisation tool first; add scikit-learn for ML; then layer on LLM APIs, a framework and a vector store when you build GenAI projects. Tools change quickly — fundamentals transfer.