Company Description Suave is a modern buy-now, pay-later platform that gives customers the flexibility to shop the items they love and pay later, both in-store and online. The company focuses on making shopping as easy as possible by allowing users to pay for items they need in easy payments.

You will own Suave's credit decisioning end to end. Today we run a Python-based scoring module alongside a deterministic payslip fraud-detection layer. It works, but it was built for a smaller company. Your job is to turn six years of data into a modern, measurable, and explainable scoring model that approves more good customers, declines fewer of them by mistake, and keeps losses low as we grow into new islands.

This is a hands-on, senior individual-contributor role with the scope to grow into leading a data team.

What you'll work withYou will build on:

- Full repayment histories since 2020, covering on-time payments, late payments, defaults, and recoveries across every installment term.
- Application and KYC data from identity verification, plus parsed payslip and income documents.
- Card transaction data showing how customers actually spend with Suave, by merchant, category, and island.
- Merchant and basket data, such as ticket size, vertical, and in-store versus online purchases.
- Repayment channel data from card and pay-by-bank payments.

- What you'll do

1. Audit the current scoring module. Measure how well it separates good and bad customers today, identify where it leaves money on the table, and document everything it currently relies on.
2. Build the data foundation. Design a clean, versioned dataset and feature store from our production systems (Laravel/PHP backend,



Python services, AWS), with clear default definitions and reliable labels.
3. Develop the next-generation scorecard. Train and validate models (for example logistic regression scorecards and gradient-boosted trees), using proper out-of-time testing and reject-inference techniques, so the model is not only learning from customers we already approved.
4. Design the decision strategy. Turn scores into approval rules, credit limits, installment-term eligibility, and pricing, then run controlled tests (champion vs. challenger) to prove impact before full rollout.
5. Put it into production. Ship the model as a reliable, low-latency service with our engineers, including monitoring for drift, approval rates, and early delinquency.
6. Keep it explainable and fair. Ensure every decision can be explained to a customer, a regulator, or a lender. Keep sensitive personal characteristics out of the model and test for unfair outcomes.
7. Support the business. Produce the portfolio and loss reporting that our lenders, funding partners, and central-bank conversations depend on.

What we're looking forMust have

- 4+ years building credit risk or underwriting models in production at a lender, BNPL, card issuer, or fintech.
- Strong Python and SQL, with hands-on experience in scikit-learn, XGBoost/LightGBM, or similar.
- A solid grasp of credit concepts such as probability of default, vintage and roll-rate analysis, Gini/KS/AUC, population stability, and reject inference.
- Experience deploying models into live systems, not only notebooks.
- The ability to explain technical trade-offs clearly to non-technical founders.

-

📌 Credit Scoring & Machine Learning Lead (Colombia)
🏢 suave
📍 Colombia