Open Source in Finance Forum London 2026

Data quality failures in financial services carry real regulatory and business risk. Traditional rule-based validation catches known issues but misses subtle patterns that emerge over time in production systems.
I built an open source real-time data quality monitoring system processing 332,000+ orders using Apache Kafka, Python, and a 3-model ML ensemble — Isolation Forest, LSTM, and Autoencoder — with SHAP explainability making every alert actionable. The system runs continuously with sub-10ms latency and automatic drift detection.
This talk covers practical lessons for financial services technologists: how to layer ML anomaly detection on top of rules-based validation, how to maintain model accuracy over time, and how open source tooling can replace expensive enterprise data quality products — saving organisations £100K+ annually.