What COVID did to our forecasting models (and what we built to handle the next shock) How Airbnb built forecasting models resilient enough to survive a global pandemic and whatever shock comes next.

What COVID did to our forecasting models (and what we built to handle the next shock)
medium.com/airbnb-engin...

A Bayesian Dirichlet autoregressive conditional heteroskedasticity model for forecasting currency shares In marketplace finance, the daily mix of billing currencies is compositional data that drive forecasting, reporting, and treasury risk. We study Airbn…

New paper out in the International Journal of Forecasting! B-DARCH is a Bayesian compositional volatility model for FX forecasting, developed at Airbnb with Rob Weiss (UCLA). A small glimpse into how we think about forecasting there.

www.sciencedirect.com/science/arti...

It’s called mars bar and they do open mic standup on Wednesdays

Directional-Shift Dirichlet ARMA Models for Compositional Time Series with Structural Break Intervention Compositional time series, vectors of proportions summing to unity observed over time, frequently exhibit structural breaks due to external shocks, policy changes, or market disruptions. Standard meth...

New preprint: Bayesian Dirichlet ARMA with directional-shift intervention for structural breaks in compositional time series. Applied to #Airbnb fee recognition during COVID.
arxiv.org/abs/2601.16821

Distributional Fitting and Tail Analysis of Lead-Time Compositions: Nights vs. Revenue on Airbnb We analyze daily lead-time distributions for two Airbnb demand metrics, Nights Booked (volume) and Gross Booking Value (revenue), treating each day's allocation across 0-365 days as a compositional ve...

I spend a lot of my time at work thinking about lead times and one of the more interesting patterns is how differently bookings, trip length, & price behave as you move from far-out reservations to last-minute stays. New paper is really just structured curiosity about this:
arxiv.org/abs/2601.12175

Staff Data Scientist, Forecasting-Finance - Careers at Airbnb Airbnb was born in 2007 when two hosts welcomed three guests to their San Francisco …

Staff DS role open on my team at Airbnb. Revenue forecasting systems: model selection, pipelines, automation. Work that actually gets used and scrutinized. Good mix of research and applications! #airbnb

careers.airbnb.com/positions/75...

Same experience with #Ox. Had to install it to dig into a related paper's code and felt like software archaeology.

Is Claude Code useful? Yes. Did it remove code instead of testing it when I prompted to write tests to increase coverage? Also yes