- Built "Marketing AI Studio", a multi-agent AI system using Google ADK to autonomously generate marketing content (emails, social posts, blogs, and articles), saving $ 1.7 million annually by eliminating third-party dependencies. Developed with FastAPI and Langchain for orchestrating agent workflows and API endpoints, Docker for scalable containerized deployment and Apigee for secure API management. Utilized PostgreSQL with checkpointers to persist agent memory across sessions and integrated Galileo for telemetry and performance logging.

​

- Developed “Scribe AI,” a FastAPI-based application integrated into Salesforce CRM that enables Verizon sales representatives to generate high-quality outreach emails using Gemini Pro 2.0. Leveraged Milvus as a real-time vector database for storing and retrieving information about Verizon’s plans and promotions, reducing email writing time by 78%

- Implemented “Sales Infobot” using Dialogflow CX integrated with Google Cloud Datastore to provide Verizon sales teams with instant, accurate information on products, services, and policies, achieving 92% intent–response accuracy and reducing query resolution time by 65%.

- Automated Q&A testing framework for Sales Infobot by developing a Gemini-powered system to generate question-answer pairs from Verizon's knowledge base. Implemented evaluation pipelines with metrics including relevancy scoring, groundedness assessment, and contextual accuracy, resulting in 85% reduction in manual testing effort