All work
LLM & Agents · Enghouse Interactive · 2024
Automated LLM Evaluation Framework
Automated quality assessment for LLM-generated output, plus unit and E2E tests for LLM services.
01 The problem
LLM output quality was judged manually and inconsistently, slowing every model or prompt change.
02 What I did
- Designed scoring for summarisation, extraction and sentiment tasks across model families.
- Wired evaluation into unit / E2E tests and Jenkins CI/CD.
03 Impact
- Model and prompt changes validated automatically
- Faster, safer iteration on LLM services
Client work — details generalised to respect confidentiality. Happy to discuss specifics in an interview.