Building with agents 0/15

Building with agents

AI in production

The cost, latency and failure handling of an AI feature once people depend on it: caching, model choice, fallbacks and monitoring. Token prices and response times decide which designs are worth shipping.

Reading

WhiteBox lessons1

Papers and essays1

  1. Patterns for Building LLM-based Systems & Products(opens in a new tab)

    Yan 2023 Essay 90 min

    Seven patterns for putting language models into products: evals, retrieval-augmented generation, fine-tuning, caching, guardrails, defensive UX and collecting user feedback, each explained with the papers and production examples behind it.

People to follow1

  • Chip Huyen

    Author of AI Engineering

    Writes about AI system design; her book AI Engineering covers building applications on top of foundation models.