All work

IHS Markit2013 — 2016

NLP Email Classification

A deep-learning system that reads and routes operational email automatically — built from zero and served as a web service to global operations teams.

  • 18 hrsmanual effort saved per day
  • Globaloperations rollout

Architecture · simplified

  1. Inbound emailHigh-volume operational mailboxes
  2. NLP modelDeep-learning classifier
  3. Web serviceClassification API
  4. RoutingRight queue, right team

Context

Global operations teams spent hours every day reading and sorting incoming email before any real work could start.

The challenge

Build something from nothing: prove that NLP could reliably understand operational email, then turn a prototype into a service teams would trust with their daily workflow.

What I did

  • Led the 0 → 1 initiative as an innovation SME, from identifying the opportunity to rollout.
  • Built a deep-learning email classifier and served it as a web service so it could plug into existing operational workflows.
  • Helped teams adopt it, alongside broader automation and analytics initiatives.

Outcome

The system removed 18 hours of manual effort every day across global operations — and became the first of many automation and AI projects.

↑↓ move↵ open