At Crayon Data, deployed to Abu Dhabi Islamic Bank, I built and deployed home finance and personal finance propensity models. Campaigns targeting customers selected by these models converted at 4x the rate of earlier rule-based campaigns.
I also built a liability drop-churn model for retention outreach and an employer-name deduplication pipeline that consolidated about 400,000 variants into roughly 10,000 canonical entities. Iām now building a Finance Control agent system using LangGraph, with human-in-the-loop approvals and safeguards against hallucination.
At HSBC, I designed an end-to-end process data model and independently built and productionized a Python/PM4Py process-mining application; root-cause analysis surfaced 2,000+ hours per year of recoverable inefficiency. I also co-developed an OCR solution that cut manual effort by 80%, and earlier delivered recommendation and data-quality solutions at Latent View Analytics.
