I build decision systems for complex domains. My work turns interacting variables, real-world constraints, and incomplete evidence into choices people can understand, test, and improve.
Represent what shapes the outcome
I start with the outcome people want to improve. Then I separate what can be changed from what must be accounted for, identify the constraints that limit the choice, and make uncertainty visible. That representation works across problems that otherwise look unrelated—from resource allocation to fermentation conditions, protein search, and private document intelligence.
Simulate options and decide what to do
A useful system does more than predict. It compares realistic scenarios, respects the limits people actually face, and finds the strongest feasible intervention, allocation, or experiment. My work in forecasting, causal inference, targeting, and resource allocation grounds those recommendations in measurable outcomes.
Make decision systems interactive with AI
Interactive AI—including local language models, conversational analytics, and adaptive workflows—can give people a natural way to explore evidence and use a decision system. The language model is the interface—not the source of truth. The calculations, source evidence, privacy rules, and human approvals remain explicit.
Learn from what happens next
Every recommendation creates new evidence. I design systems to measure the result, update what is known, and improve the next decision. Scientific research is an especially demanding proving ground because experiments are expensive and uncertainty cannot be ignored; those lessons carry directly into operational systems.
I am a senior applied machine-learning and data-science leader with a PhD in Biological Sciences, specializing in molecular biology. That combination of operational experience and scientific depth lets me carry the same decision framework across very different domains without losing the details that matter.