Worked on backend stability for the cloudera/CML_AMP_RAG_Studio repository, focusing on resolving a critical initialization-order bug in the summary index component. Addressed a core reliability issue by refactoring the global settings initialization process, ensuring that LLM, embedding model, and sentence splitter configurations are set before index storage and document summary loading occur. This Python-based solution reduced startup errors and prevented failures related to improper initialization sequencing. The work leveraged backend development and LLM integration skills, laying a stable foundation for future feature development by improving the robustness of the initialization sequence without introducing new user-facing features during the period.