AI GUIDEREAD IN ORDER · EN / RU
Advanced — design and evaluate AI workflows
A reading collection on model choice, context, retrieved knowledge, multimodal workflows, agents and evaluation, with examples for further exploration.
About this collection
Explain, design and evaluate an AI workflow using appropriate model, context, retrieval, tools and human decisions.
8 articles available in this view. Read at your own pace and explore the examples that are useful to you.
Examples to explore
- An identified and permission-scoped source pack
- A completed prototype conversation
- A system diagram
- A test report with actual and predicted results kept separate
This is a personal reference collection. There is no enrolment, assessment or certificate. Use fictional information when trying examples and check results against the original sources.
Articles are available here after publication. Drafts remain private until their editorial review is complete.
8 available articles. No account or profession selection needed.
- Choose a model and decide what to change →
Distinguish pretraining, inference, prompting, retrieval and fine-tuning; design a controlled comparison for a real task.
- Give AI the right context and know where its knowledge comes from →
Understand context, sources, retrieval, memory and fine-tuning through a small fictional policy task.
- Build a knowledge workflow and test retrieval separately →
Follow documents from ingestion to a sourced answer; practise retrieval, conflicting versions and permission-aware failure handling.
- Design a workflow across text, images, audio and data →
Separate recognition, interpretation, calculation and generation, then preserve evidence through a worked learning-pack example.
- Understand tools, workflows and agents →
Distinguish tools, fixed workflows and agents, with approval and failure handling.
- Read an AI benchmark without turning it into a promise →
Understand what was measured, compare reliability and cost, and turn a headline into a small test for your own task.
- Choose a useful AI project before choosing a platform →
Compare three agency tasks, define a small pilot and count review effort as part of the result.
- Finish Advanced: design and evaluate an AI knowledge assistant →
Complete an evidence-backed prototype conversation, a workflow design, failure tests and a measured decision about what to improve.