Gioia Zheng

Python · PyTorch · Go · FAISS · evaluation pipelines · information retrieval · RAG diagnostics · reproducible ML systems.

Reproducible evaluation workflows for retrieval and RAG systems, plus RL agents for game and embodied settings.

rag-observatory and msmarco-genqa: trace-based RAG diagnostics, controlled retrieval experiments, and experiment reports.

Research-engineering conversations around IR, RAG/LLM evaluation, reproducible ML, failure analysis, and careful open-source maintenance.