All projects

Internship · Tata Teleservices · 2025

An automated way for the security team at Tata Teleservices to get through 1,000+ abuse emails a day without reading every single one.

When
Jun to Nov 2025
Where
Tata Teleservices
My role
AI/ML Intern
Built with
Python, NLP, LLM APIs, Outlook API, Azure
Azure
running in production
1,000+
abuse emails a day
95%
less manual analysis
10+ hrs
saved every week

01The problem

The security team received more than 1,000 abuse emails every day, and someone had to go through each one to decide whether it mattered. The security reports behind them were long and unstructured, so pulling out what to act on took a lot of time.

02What I built

A classifier, trained on more than 18,000 labeled emails, that reads each new email and its metadata and sorts it automatically.

An LLM-based summarization workflow that turns messy, unstructured security reports into short summaries the team can actually act on.

I connected it to Outlook through its API, so it works inside the inbox the team already uses, and deployed it to production on Azure.

03How it went

It's running in production. Manual analysis dropped by more than 95%, which saves the security team over 10 hours every week.

Next projectMultimodal breast cancer classificationA multimodal pipeline at IIT Bombay that classifies mammograms on the BI-RADS scale.