.svg)
.avif)
.avif)
AI-powered loan origination and credit scoring software for a South African FinTech
We developed a real-time credit risk optimization platform for Beam, a rising South African neo-credit bureau. This API-first loan decisioning software empowers better collaboration and transparency across credit, engineering, and executive teams.
.svg)
Business Requirements
Beam, an early-stage fintech, needed to quickly launch a credit platform for Tier 2 banks and SME lenders. With a lean team, they sought a hands-on partner to provide not just implementation, but strategic input, product vision, and long-term scalability.The core of Ignite’s business is helping retailers manage strategy, commerce, and operations through a unified, gamified platform that drives sales and service.
Solution
PYGIO partnered with Beam to define a clear roadmap and build a scalable, cloud-native lending platform. We delivered a data aggregation layer and AI-powered tools, including onboarding, credit scoring, operations, risk analytics, and collections. Our ML expertise supported predictive scorecards and personalized recommendations.We stepped in to enhance the workflow tracking feature that enabled the alignment of business and sales representatives.
80%+
reduction in analyst time across affordability and behavioural risk workflows
87%
alignment between Beam’s AI-powered CashScore and human underwriters across 1,200+ cases
24%
uplift in segmentation accuracy cut bad debt to under 4% for thin-file clients
<60s
response latency with 100% uptime, enabled by containerized credit scoring APIs
We designed and built a fintech software platform that shares real-time, high-quality, consumer-permissioned data that creates stronger connections between lenders and borrowers.
Beam connects directly to applicants' bank accounts to access real-time financial data, rather than relying on potentially manipulated PDF statements. This direct connection provides:
Beam connects directly to applicants' bank accounts to access real-time financial data, rather than relying on potentially manipulated PDF statements. This direct connection provides:
- Protection against fraud and manual tampering
- Real-time data transmission to lenders
- 100% transparency in the lending process
- Access to a "truth file" of users' actual spending behaviour and financial position
To address common ML challenges, we employed several techniques:
- Data augmentation and transfer learning to combat data bias and overfitting
- Diversified data sources to ensure models are trained on datasets representing wide-ranging demographics and financial behaviors
- Focus on building larger, more comprehensive datasets
- Emphasis on creating more inclusive and fair AI models
The core principle is that users voluntarily grant permission for the most transparent possible analysis of their financial position through direct bank data access.
Traditional credit scoring creates a "chicken-and-egg problem" - people need existing credit to be scored, but can't get credit without a score. PYGIO solved this by:
- Using bank statement data instead of requiring credit history
- Serving "thin-file customers" ( ~30% of South Africans)
- Leveraging that ~90% of South Africans have bank accounts vs. limited credit histories
The approach we chose for Beam:
- Analyses bank transaction data at a granular level for richer insights
- Provides real-time, up-to-date financial behavior analysis
- Can re-score users frequently (hourly, daily, weekly, or monthly)
- Processes data to the very last transaction, even if just 30 minutes old
Technical implementation highlights:
- Built using open-source machine learning architecture
- Proprietary approaches to data sourcing and processing
- Seamless integration into Beam’s full-stack lending platform
- Merged traditional and alternative credit data for superior accuracy
Tech Stack we work with
Gallery
Timothy Strike
Head of Product at Beam
PYGIO accelerated our roadmap by at least 6 months. Their team didn’t just build ML models – they built an engine that transformed our data into actionable, professional-grade insights. From a product and engineering perspective, they helped us leapfrog years of experimentation.
Client
Beam
Industry
Financial Services
Project start
2022 - 2025
Country
South Africa
Project focus
Project Management
Software Development
Data Science
Software Development
Data Science
Services
#Product Development
#Project Management
#Product Design
#DataScience
#Project Management
#Product Design
#DataScience