At Cashea, my work centers on how engineering teams use large language models in production. I build tools to test prompt changes, monitor model behavior, and trace responses in credit-risk applications.
Before Cashea, I led a small engineering team at BuildWiseCity. We turned daily construction reports into dashboards that let project teams trace a figure back to the original document. On a separate financial-risk platform, our score rose from 34% to 94% on an internal evaluation of extraction and calculations from Arabic and English audited statements.
My portfolio also covers model inference. In one experiment, I ran GLM-5.2 on eight Blackwell GPUs and verified that it could retrieve a value from a 240,000-token prompt.
Open to remote senior and lead individual-contributor roles in applied AI, backend systems, and forward-deployed engineering. Miami, FL (ET). English and Spanish.