I solve large-scale business problems with market-level forecasting solutions for real estate market dynamics from the US to ZIP code level totaling close to 100,000 regions.In collaboration with other Applied Scientists, (T)PMs, and my business stakeholders, I build a large variety of forecasting products serving several key areas of the business. Examples of applications include: Working on the portfolio (risk) management and optimization for Zillow Offers, the Principal Agent's revenue and pricing optimization, and optimizing Marketing campaigns as a function of the expected market environments faced by individual consumers.For each application, it is important to collaborate with technical and business stakeholders to decompose the business problem into forecasting solutions in the time-dimension and distributional solutions in the cross-sectional dimension.This involves navigating technical and organizational challenges, take uncertain project requirements, map them to technical assumptions that need to be coordinated across technical teams, develop candidate solutions, drive acceptance across business & technical stakeholders, and finally bring the agreed solution from end to end into production.To drive trust and adoption of my work, I employ state-of-the-art model & output validation with graphics and explanations for stakeholders at scale. To ensure robust solutions for the business problem, I estimate parameters using block-wise bootstrapping, tune hyperparameters with proper cross-validation, and provide ample evidence of model performance by leveraging cutting-edge backtesting frameworks.

I took an existing Forecasting environment where Forecasting projects were done in R, in a hard-to-maintain codebase, scheduled by Microsoft Task Scheduler, with little model validation, output validation, or visibility for stakeholders and transformed it into Zillow's first-of-its-kind AWS-integrated and EKS-hosted Forecasting platform scheduled by Airflow with dependency management for each job running in production.As part of the platform, I developed extensive model validation and output validation techniques as well as a state-of-the-art backtesting framework in order to drive stakeholder buy-in and provide visibility into design decisions at the research stage of all forecasting projects.Further, the platform consists of a forecasting data warehouse with point-in-time revised series as well as time-safe series for proper backtesting - a first of its kind at Zillow. This enabled others to contribute to the existing universe of metrics for causal, or predictive modeling, and had large externalities for the productivity of users across Zillow's Economic Research team but also got leveraged by a variety of other Machine Learning teams at Zillow Group.Today, the whole platform is used by 15 other Applied Scientists across 3 teams researching solutions for causal & predictive ML problems and shipping them into production in a self-serving manner with limited need for daily Engineering support.Using the newly built end-to-end platform, Applied Scientists are able to bring down production time from ingestion to publication by -78% (from 14 business days to 3 business days).Note: Economist - Forecasting belongs to the Applied Science job family and is a change in name only.

For Zillow's Economic Research team I built causal Machine Learning models estimating structural parameters and price premiums of specific home features across 130M homes while exposing geographical as well as home-segment heterogeneity in treatment effects.Became the first member of Zillow's Forecasting Team where I led initial design rounds with the goal of building accurate, timeline, and explainable forecasts for 100,000 regions by building Zillow's fully automated Forecasting Platform. I collaborated with Engineering and downstream stakeholders to understand the use-cases and features needed to adopt best-in-class Forecasting techniques at ZG and get buy-in from stakeholders regarding the platform's output.

For Zillow's Economic Research Team I've developed metrics & indices from large upstream tables capturing market dynamics by reducing dimensionality into measures that were easier to interpret by humans and machines.My work has been used internally for market classification and has been published externally to turn on the lights for consumers about their local housing market dynamics.

Working with Prof. Dr. Matthias Neuenkirch, Chair for Empirical Research in Economics;Research on the effect of the Brexit vote on the economic performance of the UK;Use of Factor Methods, Bayesian Estimation, Matching, Entropy Balancing in R and Matlab;Working with Prof. Dr. Christian Bauer, Chair of Monetary Economics;Giving lectures as the teaching assistant in Microeconomics (undergraduate level).Conception of graduate-level lecture: International Macroeconomics.

Working with Prof. Dr. Christian Bauer, Chair of Monetary Economics;Using statistical packages such as R, Stata, and Matlab as a research assistant;Teaching assistant in Microeconomics, Scientific Working and Writing (undergraduate level).Working with Prof. Xenia Matschke, PhD, Chair for International Economic Policy;Using Stata to assist with data management and regression analysis.

Working with Prof. Dr. Mueller-Fuerstenberger, Chair for Local and Environmental Economics;Teaching Assistant in Macroeconomics (undergraduate level).Working with Prof. Xenia Matschke, PhD, Chair for International Economic Policy as her Research Assistant.Working with Prof. Dr. Walter Schertler, Chair for Organization and Strategic Management;Preparing lectures for Intermediate Business Administration I, Organisation and Strategic Management;Giving lectures in Scientific Working and Writing (undergraduate level).