Statistical Machine Learning Researcher

Traditional credit was built for people who already have money. Requirements for credit history, collateral, and costly underwriting create insurmountable barriers for those who need capital most. Over 1.4 billion people lack access to credit. A vendor in Lagos earns cash daily but can't prove a steady income. A Colombian nurse with years of perfect informal repayments remains invisible to banks.

We built an alternative called Credit. Since December 2024, it has issued hundreds of thousands of undercollateralized loans using stablecoins. People from around the world have used these loans to pay for things like groceries, medicine, and transportation. Backed by $6.6 million from Paradigm and Nascent, we're scaling a system that has already reached half a million unique borrowers. Help us take it to the next level.

About the role

We seek a talented individual to build and improve the adaptive decision systems behind Credit, our leading undercollateralized lending system. You'll design models that learn borrower behavior over time and make optimal lending decisions under uncertainty, balancing exploration with exploitation across hundreds of thousands of users worldwide. You'll also develop the offline evaluation and monitoring infrastructure to safely validate these systems before deployment.

Stack

Python

TypeScript

Bayesian/probabilistic modeling (PyMC, Stan, NumPyro, or similar)

Bandit and RL frameworks

SQL

Grafana/PrometheusKey responsibilities

Design, maintain, and optimize adaptive credit policies using methods like Thompson Sampling, contextual bandits, and Bayesian models

Formulate lending decisions as sequential decision problems under uncertainty (e.g., progressive trust-building, dynamic credit limits, risk-aware exploration)

Build offline evaluation frameworks to safely test new policies before going live

Model borrower behavior with limited, non-stationary data across diverse emerging-market populations

Develop tools, alerts, and analytics to monitor policy performance and detect distribution shifts

Collaborate with engineering to implement decision systems in productionRequirements

Graduate degree in a quantitative field such as mathematics, physics, or computer science.

Very strong foundations in probabilistic modeling and Bayesian inference

Experience applying bandit algorithms to real-world decision problems in production (credit, pricing, recommendations, resource allocation, or similar)

Ability to make and defend pragmatic tradeoffs (e.g., heuristic > learned policy, simple bandit > deep RL) based on empirical evidence and to communicate them well verbally and in internal research write-ups.

Experience in Python, Typescript, SQL, and programming for data analysis

Exceptional problem-solving skills and attention to detailNice to have

Experience in traditional credit, lending, fintech, or insurance, especially in emerging markets or data-scarce environments

Published work or open-source contributions in bandits, Bayesian ML, or sequential decision-making

Experience with DeFi protocols, especially lending or credit systems

Familiarity with blockchain data indexing and onchain analytics

Divine Research is an equal opportunity employer.