Airbnb Data Scientist Interview Questions

As Airbnb implements machine learning models for various pricing and recommendation systems, a robust model versioning system is essential for maintaining and deploying different iterations effectively.
Problem Statement:
You need to design a class ModelVersionRegistry that keeps track of different versions of models with distinct identifiers. Each model could be associated with metadata such as creation date, performance metrics, etc. Implement the class with the following methods:
- def add_model(identifier: str, metadata: dict) -> None: Registers a new model with given identifier and its associated metadata.
- def get_metadata(identifier: str) -> dict: Returns the metadata associated with the specified identifier.
- def find_best_model() -> str: Returns the identifier of the model with the best performance based on some predefined metric stored in the metadata.Example 1:
Input: registry = ModelVersionRegistry()
registry.add_model('v1', {'accuracy': 0.85, 'created': '2023-01-01'})
registry.get_metadata('v1')
Output: {'accuracy': 0.85, 'created': '2023-01-01'}Constraints:
- Model identifiers are unique and will be alphanumeric strings.
- Metadata will contain various performance metrics.

Airbnb often analyzes visitor trends over time using tree structures to represent daily bookings across different regions. Efficient traversal methods help understand user engagement better.
Problem Statement:
Given a binary tree representing bookings, where each node contains a visitor_count, return the maximum sum of visitors from the root to any leaf. You should implement the function:
- def max_visitor_path(root: TreeNode) -> int:
This function should return the maximum sum of visitor_count along any path from the root to a leaf node.Example 1:
Input: root = [2, 7, 5, 3, 4, None, 1]
Output: 12
Explanation: The path with the maximum sum is 2 -> 7 -> 3.Constraints:
- 1 <= number of nodes <= 1000
- 0 <= visitor_count <= 10000

As Airbnb improves its pricing strategy, it needs to make real-time pricing decisions based on demand fluctuations in various cities.
Problem Statement:
You are given a list of potential base_prices for a rental property over n days and the rental price can be adjusted once for each given day to either increase or decrease by a percentage (given as a list of percent_changes). Your task is to maximize the price for the last day. You should implement the function:
- def max_pricing(base_prices: List[int], percent_changes: List[int]) -> int:
This function should return the maximum listing price possible on day n-1 by applying one change at most to each previous day’s price.Example 1:
Input: base_prices = [100, 200, 300], percent_changes = [10, -20, 30]
Output: 390
Explanation: Adjust the price on the second day to 240 (decrease of 20%) and increase the final day price by 30%, yielding 390.Constraints:
- 1 <= n <= 10^4
- 0 <= base_prices[i] <= 10^6
- -100 <= percent_changes[i] <= 100

Airbnb continually aims to improve the user experience by personalizing search results based on user interactions. Using a sliding window approach can help identify the most relevant listings based on user preferences and browsing history.
Problem Statement:
Given a list of n listings where each listing has a relevance_score and a user’s preference defined as a window_size, return the maximum total relevance_score the user can see in any contiguous subarray of size window_size. You should implement the function:
- def max_relevance_score(listings: List[int], window_size: int) -> int:
This function should return the maximum sum of relevance_scores of any subarray of size window_size.Example 1:
Input: listings = [1, 3, 2, 5, 4], window_size = 3
Output: 10
Explanation: The sum of relevance scores from indices 1 to 3 (3 + 2 + 5) is the maximum.Constraints:
- 1 <= n <= 10^5
- 0 <= listings[i] <= 10^6
- 1 <= window_size <= n