Description
Discover how Novacomp transformed the Machine Learning infrastructure of a major Fintech in the banking sector, resolving visibility challenges for in-production predictions and optimizing the model lifecycle. Through a microservices-based architecture, event-driven monitoring, and real-time experimentation frameworks, the solution standardized deployments and maximized the operational performance of the embedded lending system.
What will you discover in this document?
By reading this case study, you will learn:
- The challenge of Machine Learning monitoring and deployment: The lack of visibility into prediction behavior within production environments, the absence of infrastructure to execute A/B testing, and the lack of standardization across deployment processes.
- The integration approach and technical collaboration: The inclusion of talent under a Staff Augmentation model working in synergy with internal teams, actively participating in agile ceremonies, and collaborating with Data Engineering and Analytics cells.
- The technological ecosystem and MLOps architecture: The implementation of an advanced stack composed of Python, FastAPI, Flask, Docker, Apache Pulsar, Split.io, and Weights & Biases, integrated with predictive modeling libraries such as XGBoost, CatBoost, and Scikit-learn.
- Operational results and technical autonomy: The development of an event-driven real-time monitoring system, a significant reduction in latency through asynchronous code, clear documentation for autonomous deployments, and the enablement of parallel A/B testing.