Beyond LLM-as-a-Judge: Deterministic Metrics for Multilingual Generative Text Evaluation

Ship a Python package for deterministic model evaluation. This replaces expensive LLM-as-a-judge setups with fast, sub-1B parameter models for CI/CD.

Suggested repo: omniscore-lite

"Replace your $100/day LLM judge with a local, deterministic metric."

Estimated effort: 50h