Projects
Verta is an AI-powered shopping assistant designed to enhance customer experiences by answering product-related queries with precision and relevance. Leveraging user reviews and product descriptions, Verta delivers honest, concise, and insightful responses that empower users to make informed purchasing decisions.
Technology:
RAG, MultiAgent, Groq, HuggingFace, LangGraph, LangFuse, Docker, CI/CD, Git Actions, Google Cloud Run, FastAPI
Timeline:
Sep 2024 - Dec 2024
Key Achievements:
Multi-Agent RAG Chatbot capable of answer questions related to product metadata and user-reviews
Achieved exceptional Context Recall and Context Relevancy scores of 0.95, ensuring responses are accurate and contextually relevant, evaluated using WandB Weave.
Reduced query response times by 3x through advanced state management and cache memory utilization, enabling seamless interaction even under high user load.
Built on FastAPI and deployed on Google Cloud Run to support high-volume queries with efficiency and scalability.
