← thecodex.expert · The Codex Family of Knowledge
Learning Path

🗄️ Data, AI, and Backend Services

Data science, machine learning, REST APIs, microservices — the server-side work that powers every application. Python and SQL are the entry point to almost all of it.

1

Python — the universal data language

Why first: Python is the dominant language for data science, machine learning, and scripting. NumPy, Pandas, scikit-learn, PyTorch, TensorFlow — all Python. FastAPI and Django are strong backend frameworks. Python is the most hirable language right now, covering more job descriptions than any other.

What to learn first: Python fundamentals (lists, dicts, functions, classes, comprehensions) → file and JSON handling → HTTP requests with httpx/requests → async/await → type hints with mypy. Don't jump to frameworks until you're comfortable with the language itself.

For data science specifically: After core Python, learn NumPy (arrays and vectorised operations), Pandas (data manipulation), Matplotlib/Seaborn (visualisation), and scikit-learn (machine learning). These four cover the vast majority of data science work.

For backend APIs specifically: After core Python, learn FastAPI (async, typed, fast) or Django (full-featured, batteries included). FastAPI is recommended for new projects — it's async by default and produces automatic API documentation.

Time estimate: 2–3 months for solid Python fundamentals. You're ready to move on when you can write a script that fetches data from an API, transforms it, and writes it to a file.
2

SQL — every data job requires it

Why SQL is non-negotiable: Whether you're a data scientist running analyses, a backend engineer querying a database, or a data engineer building pipelines — you will write SQL. Almost all structured data lives in relational databases. SQL has been the language for querying them for 50 years and shows no signs of replacement.

What to learn: SELECT with JOINs and GROUP BY first. Then window functions (ROW_NUMBER, RANK, LAG) — these separate data analysts from people who can only write basic queries. Then CTEs for readable complex queries. Then indexes and EXPLAIN ANALYZE for performance.

Use PostgreSQL: It's free, open source, and the most feature-rich SQL database. Everything you learn in PostgreSQL transfers to MySQL, SQLite, and cloud databases (BigQuery, Redshift, Snowflake all use SQL dialects).

Time estimate: 4–6 weeks to be productive. Python + SQL makes you employable in data engineering, data analysis, and backend roles.
3

Scale: Java, Go, or deeper Python

Java if you're going into large enterprise engineering teams, working on Android, or building the kind of high-throughput backend services that run at scale. Java's ecosystem (Spring Boot, Kafka, Hadoop, Spark) dominates enterprise data engineering. Virtual threads (Java 21) make Java competitive with Go for high-concurrency services.

Go if you're building microservices, cloud-native applications, or developer tools. Go's goroutines, channel-based concurrency, and single-binary deployment make it excellent for building reliable services that handle high traffic with minimal resource usage.

Deeper Python if you're focused on data science, ML engineering, or AI. Python 3.12+ is significantly faster than earlier versions. Learning async Python, typing, and the ML ecosystem (PyTorch, JAX, Hugging Face) can take years to master fully.

The AI/ML engineering shortcut

If your goal is specifically AI and ML engineering (not data analysis), the fastest path is: Python → NumPy → PyTorch → Hugging Face. Everything else you'll pick up along the way. The maths (linear algebra, calculus, probability) matters more than knowing multiple languages in this domain.