Worked on the Canner/WrenAI repository to enhance the reliability of DeepSeek API integrations by developing a robust JSON extraction and validation utility in Python. This solution focused on parsing JSON content from Markdown code blocks, ensuring that only valid JSON was processed by downstream services. By introducing rigorous error handling and backend validation logic, the work addressed issues with invalid JSON responses from DeepSeek, reducing parsing errors and improving the stability of AI-driven features. The approach emphasized traceability through well-documented commits and aligned with best practices in backend development, API integration, and error handling to strengthen data integrity.