At DM Financial NL, I led the design of an end-to-end seven-agent pipeline for insurance claims, covering intake, document parsing, coverage checks, fraud screening, damage assessment, decisions, and compliance. I also implemented observability and audit dashboards for agent workflows.
At ING Netherlands, I developed and validated forward-looking credit risk models across retail and wholesale portfolios, using statistical learning, time-series analysis, Python, and SQL. I worked directly with ING Audit and ECB stakeholders on model risk, validation, governance, and compliance.
At ING Netherlands, I built a financial document assistant that retrieved answers across 10-K filings, earnings calls, and internal policies. I also implemented agent evaluation and adaptive memory, with the memory framework improving recall by approximately 35%.
My work also includes a GRPO/LoRA training pipeline for Qwen coding agents, an experience-driven trading agent, and a multi-agent model risk assistant. I focus on reliable, explainable AI, from predictive models to autonomous decision-making.
