Philipp Eisenhauer
Science alone doesn't create value. Productionizing, automating, and scaling it does.
I build production systems that translate scientific models into decision frameworks — so organizations can act on their evidence at scale.
Science × Engineering × Decision = Impact
Science informs what to build — measurement and modeling that identifies what drove outcomes and why. Engineering turns that into a reliable, repeatable process at scale. Decision builds on both — acting on evidence at the right moment, deferring when uncertainty is too high, and closing the loop on actual versus predicted outcomes with each new cycle.
Showcase
Structural Models for Policy-Making
Policy recommendations depend on model parameters that are never known exactly. This paper combines robust optimization with statistical decision theory to quantify how that uncertainty propagates — and what it means for the decisions you'd actually make. Conditionally accepted at the International Economic Review.
Read the paper →Impact Engine
Measuring impact, judging evidence, and allocating resources are usually separate, manual workflows. Impact Engine is an open-source Python ecosystem that connects all three — independent packages that measure impact, evaluate evidence, and allocate resources, wired together by an orchestrator at runtime.
Explore the Impact Engine →Inside Amazon's AI Factory
AI investments are expensive and their returns are hard to measure. This article documents the causal measurement framework behind Amazon's Catalog AI — randomized experiments that quantify business value and inform production decisions at scale. Published in the Harvard Business Review.
Read the article →Experience
Amazon.com
Science lead for impact measurement platforms serving Amazon's product catalog and AI initiatives. Develop and deploy production systems for experiment design, causal inference, and return on investment (ROI) evaluation that directly inform high-stakes investment decisions.
University of Washington
Instructing a course on causal data analysis and scientific computing in the Department of Economics, cross-listed with the Computer Science and Statistics departments.
University of Bonn
Advanced decision-making under uncertainty through robust computational frameworks. Integrated statistical decision theory, robust optimization, and econometric modeling to develop tools that quantify and navigate uncertainty in dynamic systems.
Publications
Business
- Philipp Eisenhauer (2025). Science-backed Decisions for High Stakes Investments Under Uncertainty. Working Paper.
- Philipp Eisenhauer, Stefan Thomke, and Puneet Sahni (2025). Inside Amazon's AI Factory. Harvard Business Review.
- Abraham Asfaw, Philipp Eisenhauer, and Andrea Scarinci (2024). Evaluating the Helpfulness of AI-enhanced Catalogue Data. Amazon Science Blog.
Academia
- Philipp Eisenhauer, Janoś Gabler, Lena Janys, and Christopher Walsh (2025). Structural Models for Policy-Making: Coping with Parametric Uncertainty. International Economic Review (conditionally accepted).
- Manudeep Bhuller, Philipp Eisenhauer, and Moritz Mendel (2025). Sequential Choices, Option Values, and the Returns to Education. Quantitative Economics (2nd resubmission).
- Maximilian Blesch and Philipp Eisenhauer (2021). Robust Decision-Making Under Risk and Ambiguity. arXiv.
- Philipp Eisenhauer, James J. Heckman, and Edward Vytlacil (2015). The Generalized Roy Model and the Cost-Benefit Analysis of Social Programs. Journal of Political Economy, 123(2), 413–443.
- Philipp Eisenhauer, James J. Heckman, and Stefano Mosso (2015). Estimation of Dynamic Discrete Choice Models by Maximum Likelihood and the Simulated Method of Moments. International Economic Review, 56(2), 331–357.
Teaching
- Business Decisions: Teaches students to connect data insights to organizational action through systematic impact measurement and resource allocation. Covers causal inference for business, experimentation design, and building decision systems that scale.
- Data Science: Develops the ability to make and critically evaluate causal claims using real-world data. Covers research design, program evaluation, and microeconometric methods with hands-on application in Python.
- Scientific Computing: Builds production-grade computational skills for solving complex economic and statistical problems. Covers numerical optimization, software engineering practices, and high-performance computing in cloud environments.