Skip to content

Revise — the last-minute layer

~40 min total · 33 concepts in 60-second sheets · built for the last two days

One concept per sheet: what it is, the trap, and a practice question that tests it. Skim the category you're weakest in, or read all six in one sitting.

How to use these

  • 2 days out: read all six categories (~40 min), ticking off concepts you know cold
  • 1 day out: re-read only the ones you ticked as shaky
  • Last hour: the cheat sheet — every must-remember on one printable page

The six categories

  • 🌐 Web & APIs

    CORS · HTTP status codes · caching (TTL, Cache-Control, cache-buster) · API gateway · HTTPS

    ~6 min · 5 concepts

  • 🐳 Docker & deployment

    EXPOSE vs -p · missing install · parity · layer caching · serverless limits · GitHub Pages

    ~7 min · 6 concepts

  • 🤖 LLMs & prompting

    System/user prompts · Structured Outputs · prompt injection · OWASP LLM Top 10 · embeddings

    ~6 min · 5 concepts

  • 🔍 RAG & agents

    RAG pipeline · chunking · stale documents · asyncio.gather · agent loop · MCP · sandboxing

    ~7 min · 6 concepts

  • 📊 Data & ML

    Parquet vs text formats · data leakage · pandas operations · ETL patterns · finetuning ladder · quantization

    ~7 min · 6 concepts

  • 🔧 Git, security & practices

    Git daily loop · merge conflicts · secrets workflow · race conditions · dependency management

    ~6 min · 5 concepts

The final hour

  • 📋 Exam-day cheat sheet

    Every must-remember, trap, and wrong belief on one page. The marks-ramp time budget. The task-to-tool decision table. Print it and read it before you walk in.

New topics are here at full weight

W4 (RAG), W5 (Agents), and W8 (ML platforms) have no PYQ history because they're new this term — not because they're less important. The GAs test them heavily and the exam will too. They're treated equally throughout.