No. 005
Paints Receivables ML Model
DataAI
Seen/ Management Intern, Digital and Finance ยท Apr 2024 to May 2024
Project entry
_
Moves
Model the Risk
Built an ML model predicting accounts-receivable risk and per-dealer default probability from 2+ years of dealer data.
Validate It
Validated the model at about 93% accuracy before it went near a real credit decision.
Design the Framework
Built a risk-based dynamic credit-limit framework segmenting dealers by payment behaviour and credit risk.
Pitch It Up
Presented findings and the framework to senior finance and digital leadership.
Outcome
Projected to cut overdue accounts by roughly 25%, recognized by the CDO, and led to a Pre-Placement Offer.