LLM engineering

Engineering LLM applications for production: token costs, prompt caching, evaluation, and security against prompt injection.

13Articles

8Topics

1Author

Prompt injection tops OWASP's list of LLM risks because models cannot tell instructions from data. Detection helps a little; architecture is what…

M morpheus 9 min

Most LLM bills are dominated by the same tokens sent over and over. Caching, batching, routing and shorter outputs, with the arithmetic…

M morpheus 8 min