skip to content
Ayush Kedia | Portfolio
Table of Contents

View GitHub Repository

Project Overview

Architected an end-to-end NLP pipeline analyzing 1,000+ fintech and retail support leads to identify sales funnel drop-offs and turnaround time (TAT) bottlenecks. This system automates intent extraction to categorize complex user queries into actionable financial insights.

Key Insights & Impact

Data Engineering & NLP: Cleaned raw datasets and engineered custom metrics using Python (Pandas) and Google Sheets.

Intent Classification: Leveraged Hugging Face Transformers (BART) and Llama Embedders to categorize Net Promoter Scores (NPS) and user sentiment.

Operational Analytics: Calculated stage-to-stage conversion rates to pinpoint inefficiencies in the support funnel.

Visualization: Integrated data with Looker Studio to provide real-time visibility into customer sentiment and operational performance.