Lyft Data Scientist Intern Interview Experience — Open-Ended Ride Request Demand Forecasting Case

This was for the algo ML intern role.

Question: Ride Request Demand Prediction

You are given the past 5 minutes of data from the Lyft app, including:

• Sessions: the number of sessions in which the Lyft app was opened
• Requests: the number of ride requests users made
• Geohash: the city is divided into many geohash grids, each roughly 1.2 km × 600 m
• Premium multipliers: 0%, 15%, 50%, 70%, 100%

Goal: for each geohash, predict how many ride requests Lyft will receive in the next minute under each different premium multiplier.

In other words, you need to predict the next-minute request count for every geohash × 5 premium multipliers.

Overall it is a fairly open-ended ML system design / demand forecasting case. You have to pick a model, and also explain your modeling approach in light of the business scenario.