$15M+ in quality losses, found in data Cummins already had.
I led analytics on Axion’s Cummins account. Engine-failure signals were spread across warranty claims, repairs and technician notes. I built the workflows that connected them into answers engineering could act on.
Problem
Engineers were building pivot tables by hand while IoT logs, technician notes and service requests sat on servers. Standard BI tools had hit their limits on data this inconsistent.
Insight
The losses weren’t hidden. They were fragmented.
A fault only becomes a pattern once it’s linked across claims, repairs and escalations. The fix was a better join, not a better dashboard.
What I did
Built the workflows that link failures across sources, used LLMs to turn free-text failure descriptions into consistent codes, and led the team that delivered it.
Outcome
$15M+ in potential quality losses put in front of engineering. The account grew from a $60k pilot to $3M in annual revenue.