Acadbyte
Start free

GenAI in Practice

Building things that hold up outside the demo

Prompt patterns, retrieval, vector search, agents, structured output, evaluation and cost — the decisions you make once the toy version works.

Beginner–Advanced68 min9 topics
0 of 9 topics

Prompting Well

  1. Prompt Patterns: Few-shot, CoT, RolesBeginnerThree examples will beat a paragraph of careful instructions almost every time. Why does showing work so much better than telling?9 min

Grounding in Your Data

  1. Retrieval-Augmented Generation (RAG)IntermediateA model has never seen your company's refund policy. So how does it answer questions about it correctly — and cite the paragraph?7 min
  2. Embeddings for SearchIntermediateHow does a search box find 'get my money back' inside a document that only ever says 'reimbursement'?7 min
  3. Vector Databases 101IntermediateSearching ten million documents in 20 milliseconds sounds impossible. It is — so what exactly did the database quietly give up?7 min

Making Models Act

  1. Agents & Tool UseAdvancedThe moment a model can do something instead of just say something, a 95%-accurate step stops being good news. Why?8 min
  2. Structured Outputs & Function CallingIntermediate'Respond only with JSON' works 299 times, then breaks your parser at 2am. What makes it impossible to break instead of merely unlikely?7 min

Shipping Responsibly

  1. Evaluating LLM OutputsAdvancedYour prompt change looks better on the three inputs you tried. How would you know if it wasn't?8 min
  2. Cost, Latency & Model SelectionIntermediateThe best model is usually the wrong default. Reordering two paragraphs of your prompt can cut the bill by an order of magnitude. Both are the same idea.7 min
  3. Guardrails & Safety BasicsIntermediateYou can't fix prompt injection by telling the model to ignore malicious instructions. The reason why is structural — and it changes how you build.8 min
All tracks