Method
Context engineering
Curating instructions, metadata and layers of context supplied to LLMs.
20 cases from 15 companies apply Context engineering. 3 were checked in an online A/B test; 8 report results in production.
Typical tasks: Answer questions over data, Make metric definitions consistent, Evaluate models, Build a data platform and 7 more. Companies: Databricks, LinkedIn, Netflix, Grab, Spotify, Khan Academy and 9 more. Often combined with: LLM-as-a-judge, Text-to-SQL, MCP, Human in the loop, Semantic layer and 44 more.
