Panav Kumar

PROJECTS

1. LLM-Powered RAG Chatbot for Insurance Firm

• Built and deployed a chatbot using Python (FastAPI) and React.js, improving agent response times by 15–20%.
• Engineered data ingestion and cleanup pipelines with Pandas and NumPy to process large Excel datasets for vector database loading.
• Achieved 80–85% retrieval accuracy with structured and unstructured data, implementing streaming responses to cut first-token latency to ~5 seconds.

2. AI-Driven Training & Retention Solutions

• Delivered solutions for global CX and telecom clients, reducing training errors by 25%, improving speed-to-competency by 25%, and boosting attach rate by 1%.
• Built multi-agent architecture for grammar checks, policy compliance, customer simulations, and guardrail validation.
• Transformed unstructured content into Neo4j graph databases and integrated with Azure AI Search for contextual retrieval and automated assessment generation.

3. Talk-to-Data Platform

• Designed a platform for a multi-billion-dollar electronics firm to query structured financial, revenue, and engineering data via natural language.
• Developed data ingestion, cleaning, and transformation pipelines in Python (Pandas, NumPy) for SQL-based querying and analysis.
• Implemented LLM-powered SQL generation and automated chart selection with Charts.js to produce optimal visualizations.

• Oversaw creation and upkeep of organization's website from ground up utilizing WordPress and associated plugins.
• Conducted competitor analysis to inform business development strategies.
• Supported the preparation of presentations for potential clients and stakeholders.