Machine Learning Engineer, Applied Research
Whatnot, New York, New York
The role
- Lead research projects across the marketplace dynamics problem area: simulation, auction and allocation mechanics, long-term objective modeling, exploration and information value, or marketplace experimentation methods
- Take ideas from hypothesis to production: literature review, prototyping, offline validation, shadow testing, and online experiments shipped through partner teams in Discovery and the Seller org
- Build models of how the marketplace behaves as a system: learned simulators that predict segment-level effects of ranking and policy changes, and surrogate models of long-term marketplace outcomes
- Model Whatnot's actual market mechanics: auction and bidding dynamics, and discovery exposure allocation as a portfolio problem, including allocation to rising sellers
- Advance how a multi-sided live marketplace evaluates changes: off-policy evaluation, switchback and interference-robust experiment designs, and variance reduction