SELECTED WORK / 03 LLM Testing Framework
Selenium × DeepEval
Bringing real QA rigor to LLM-powered features.
01 / THE IDEA
From possibility
to practice.
A bilingual testing framework: Java Selenium drives the UI and captures LLM responses, a Python FastAPI service scores them with Gemini across 7 evaluation metrics — accuracy, relevancy, coherence, hallucination, faithfulness, compliance and toxicity.
02 / INSIDE THE SYSTEM
01
7 evaluation metrics: accuracy, relevancy, coherence, hallucination, faithfulness, compliance, toxicity
02
Java Selenium → Python FastAPI bridge over HTTP
03
Dark-themed web testing console for manual runs
04
Custom compliance rules: must-contain / must-not-contain terms
05
Per-metric scores with explanations, pass/fail verdicts
CONTINUE EXPLORING / 04