About

I came to data science and ML from neurobiology. That background gave me a durable way of thinking: systems with many interacting parts, feedback loops, and the importance of measuring the right things before drawing conclusions. I am comfortable with uncertainty and with building models that are explicitly approximate—then designing pipelines and evaluations so that those approximations are visible and improvable.

I have worked on problems in healthcare (behavioral risk, clinical decision support), transportation (ETA, demand, operations), and infrastructure more broadly. I care about the full stack: data quality, feature engineering, model choice, serving, monitoring, and documentation. I prefer to own the loop from question to deployed system rather than only the model in the middle.

I am drawn to domains where models touch real people and real infrastructure: healthcare, public transit, energy, supply chain. I am interested in how we make ML systems interpretable and robust, how we evaluate them fairly across populations, and how we integrate them into existing workflows without pretending they are infallible.

I think a lot about tradeoffs—between accuracy and simplicity, between automation and human oversight, between speed to ship and long-term maintainability. I believe in writing things down: design docs, evaluation criteria, and limitations, so that the next person (or the future me) can understand why decisions were made.

I want to work on problems that have public impact. That does not mean only nonprofits or government—it means choosing roles and projects where the outcome improves something beyond a single metric or shareholder return: patient outcomes, equitable access to services, or more reliable infrastructure. I am willing to trade some upside for clarity of purpose and for teams that take ethics and rigor seriously.

Long term, I see myself continuing to build and lead technical work in applied ML—either as an IC who owns complex systems or as someone who helps shape what gets built and how it is evaluated. I am not chasing the latest architecture for its own sake; I am interested in systems that work in production and that teams can trust.