I'm a machine learning engineer based in Atlanta. I spent most of my Georgia Tech years deep in coursework, and during my MS got my first real taste of research, contributing to a publication on debugging object detection models at VLDB. That experience stuck with me. I've been in industry since, building production ML systems, but the research pull hasn't gone away.
At NCR Voyix I build the ML infrastructure that powers time-series forecasting across hundreds of restaurant locations. That has included migrating the platform from Kubeflow to Databricks, building production pipelines that handle multi-terabyte datasets with Spark, MLflow, and BigQuery, and automating model monitoring and retraining with drift detection so deployments don't require manual intervention.
I did both my BS and MS in Computer Science at Georgia Tech. My undergrad was where I fell in love with ML, systems, and the problems that live at their intersection. During my MS year I shifted more toward research, which ended up being the most formative part of my education. I also TA'd throughout, and found I got a lot out of helping people find their footing in a field I care about.
I did an AIxBio hackathon recently focused on DNA screening, which got me thinking seriously about what it means to deploy AI in high-stakes biology contexts. Part of that work involved protein language models, and I kept running into a familiar problem: nobody really knows what's happening inside these models. With text there's at least some intuition you can apply, but with biological sequences that sanity check disappears entirely. That sent me toward mechanistic interpretability, and I've been digging into it since.