Anonymize before inference
An LLM gateway that strips personal data from prompts before the model, puts it back into the answer and traces every call without keeping the text. With measurements.
// Articles
Project: LLMOps Gateway
What I built, measured and learned in production: projects told in detail, tool notes and guides, with the commands and the numbers.
Anonymize before inference
An LLM gateway that strips personal data from prompts before the model, puts it back into the answer and traces every call without keeping the text. With measurements.
- Published
- Reading
- 12 min
- Project
- LLMOps Gateway
2 articles
-
-
Presidio, in French
What Presidio needs to learn to find names, addresses, social security numbers and IBANs in French texts, without an LLM: recognizers, check digits and measurements on 100 texts.