Senior Data Scientist at Mission Lane

Senior data scientist on fraud and credit decisioning. The year I got to see what production ML actually feels like inside a consumer bank: thresholds, approvals, losses, and the operating judgment that has to sit around any of it to make it work.

After Jumpstart, I wanted to get closer to a scaled consumer-credit environment where the feedback loops were tighter and the consequences of modeling decisions were immediate. At Mission Lane I worked on fraud and credit models in exactly that kind of setting.

The job was not just to improve model performance; it was to understand how models, policy, false positives, analytics tooling, and operations fit together in production. Thresholds were not abstract numbers. They determined who got reviewed, who got approved, and who absorbed the cost when the system was wrong.

I worked closely enough to see where model quality ended and operating reality began: approval policy, underwriting tradeoffs, customer experience, and what happens when a decisioning system is wrong at scale.

That experience made me much more interested in systems that support consequential decisions under uncertainty. It is where machine learning stopped feeling like research and started feeling like operating judgment, and where I got much more skeptical of any model that could not explain itself well enough for another operator to intervene.

This was the part of my career where I got closer to live decision systems. I was working on fraud and credit decisioning in consumer fintech, where the work had to survive contact with underwriting policy, operational constraints, analytical reporting, and real customer outcomes rather than just perform well on a static evaluation set.

The important decisions were rarely purely modeling decisions. They were questions about thresholds, false positives, who absorbs the cost of being wrong, and how model outputs fit into a broader operating process with reporting, tooling, and engineering around it. That was a useful shift for me because it made the tradeoffs explicit.

Mission Lane made me much more interested in the interface between analytics and judgment. It was one of the clearest bridges between my earlier technical work and my later interest in AI systems for high-stakes workflows, especially systems that need to be traceable enough for someone else to trust and override.