My Projects

I created this project after noticing, on my first day at Caterpillar, a gap in predicting machine part failures. To save money by repairing parts before they break down instead of replacing them, I built a system to estimate failure probability over time.

Using Python and SQL, I connected to Caterpillar's Snowflake and SAP HANA databases to pull part order and working-hour data, then cleaned and merged large, multi-source datasets by standardizing formats and joining on serial numbers and dates. I applied Weibull distribution modeling (via Python's lifelines library) to estimate each part's probability of failure within a given usage threshold (e.g., 1,000 hours), then ranked parts by risk in an exportable report. This let maintenance teams prioritize high-risk parts based on data rather than guesswork — reducing downtime and delivering an estimated 400% ROI.