Developed two production-ready features across the phidatahq/phidata and LearningCircuit/local-deep-research repositories, focusing on enhancing AI agent web search and research workflows. Delivered a SofyaTools toolkit for phidatahq/phidata, enabling agents to perform web searches, extract URL content, and conduct deep research with configurable API flags and support for both markdown and JSON outputs. Integrated a Sofya-powered web-search engine into LearningCircuit/local-deep-research, supporting two-phase retrieval and secure API-key handling. Emphasized robust backend development using Python, comprehensive unit testing with pytest, and improved security practices, while ensuring thorough documentation, CI integration, and production-level configuration for maintainable, scalable deployment.