Notes from the work.
Methodology, field observations, and compliance writing. Short pieces, written by the team, on what we're finding as we build credit infrastructure for African lenders.
FinBERT, BloombergGPT & co: a survey of finance-specialised language models
A comparative, pedagogical survey of financial language models (FinBERT, BloombergGPT, FinGPT, FinMA…): architectures, training data, performance and use cases.
Cutting the cost of risk up to 40%: building an industrial credit scoring system in an African lending institution
Most financial institutions in sub-Saharan Africa still assess credit risk through the individual judgment of loan officers, without a structured quantitative framework. This article explains how to build an industrial-grade credit scoring model, which methodological choices really matter, and what business gains to expect.
WAEMU microfinance: the next wave of failures is already on the books
The sector is growing fast, but the quality of its loan books is deteriorating faster still. Between rising arrears, under-provisioning and fintech competition, the craft-based approach to risk has reached its limits. A diagnosis, and what scoring changes in practice.