Looking for a job

pavan kumar

@pavankumar48

Message

I build production ML systems for anomaly detection, fraud detection, and LLM-powered products.

India

Message

At Tata Consultancy Services, I build production machine learning systems for a major UK telecom carrier, including billing anomaly detection that improved recall from 60% to 90%+ and reduced false alarms from 57% to 8.7%.

I also built ClaimRight, a GPT-4 and RAG health-insurance appeal platform that extracts claim data from denial letters and generates legal appeal documents in under three minutes. My work spans data preprocessing, feature engineering, model evaluation, deployment, and production monitoring using Python, AWS SageMaker, FastAPI, and PostgreSQL.

Experience

Work history, roles, and key accomplishments

Current

Data Scientist

Current

Tata Consultancy Services

May 2024 - Present (2 years 3 months)

Built ML-based billing anomaly detection system for a major UK telecom carrier, improving recall from 60% to 90%+ and reducing false alarms by 48 percentage points. Developed relational fraud detection models on telecom call networks, comparing Gradient Boosting with graph features against GraphSAGE GNN.

Python LightGBM AWS Sagemaker GraphSAGE Networkx Gradient Boosting Feature Engineering MLOps PR AUC Time Series Modeling Anomaly Detection

AI-Powered Health Insurance Appeal Platform

ClaimRight

Built ClaimRight, a GPT-4 + RAG product that reads insurance denial letters, extracts 15+ claim fields, and generates legal appeal documents in under 3 minutes using 83 Ombudsman precedent cases. Achieved 90% field extraction accuracy across legal PDFs with a FastAPI backend and PostgreSQL.

GPT 4 RAG fastAPI PostgreSQL Prompt Engineering Document Intelligence