Stripe Data Scientist Interview Questions

As Stripe scales its machine learning capabilities, a versioning system to track different model iterations is crucial for accountability and reproducibility. Design a class that manages model versions.
Problem Statement: Create a ModelRegistry class with methods to add new models, retrieve, and list models by version. Ensure models can be tagged and versioned effectively.
- addModel(model: MLModel): Adds a new model with a version.
- getModel(version: str): Retrieves a model by its version.
- listModels() -> List[str]: Returns the list of all model versions.Example 1:
Input: registry = ModelRegistry()
registry.addModel(MLModel(name='TransactionClassifier', version='1.0.0'))
model = registry.getModel('1.0.0')
Output: model.name == 'TransactionClassifier'Constraints:
- Model names can be unique strings.

Stripe offers different features based on user subscriptions and needs dynamic control over which features are enabled. Implement a feature toggle using a hash map.
Problem Statement: Design a FeatureToggle class to manage active features that can be turned on or off dynamically. The class should allow adding features and checking if they're enabled.
- addFeature(name: str): Adds a feature to the toggle.
- enableFeature(name: str): Enables a feature.
- disableFeature(name: str): Disables a feature.
- isFeatureEnabled(name: str) -> bool: Returns true if the feature is active.Example 1:
Input: toggle = FeatureToggle()
toggle.addFeature('paymentProcessing')
toggle.enableFeature('paymentProcessing')
Output: True if toggle.isFeatureEnabled('paymentProcessing')Constraints:
- 1 <= name.length <= 100

In a large-scale payment processing system, Stripe requires a method to compute real-time moving averages of transaction amounts for analytics and reporting. You need to implement an efficient algorithm.
Problem Statement: Given a stream of integers representing transaction_amounts, and an integer k, return the moving average of the last k transaction amounts as a float.Example 1:
Input: transaction_amounts = [1, 10, 3, 5, 2], k = 3
Output: [4.66666, 5.0, 3.33333]
Explanation: Moving averages computed for the last 3 transactions.Constraints:
- 0 < k <= transaction_amounts.length <= 10^5
- 0 <= transaction_amounts[i] <= 10^4

Stripe needs to predict potential fraudulent activities using historical transaction data. You need to create a machine learning model that forecasts transaction anomalies based on past trends.
Problem Statement: Given a time series of transaction_amounts, your task is to build a forecasting model that outputs potential anomalies in the next n transactions. The model should be evaluated based on precision and recall metrics.Example 1:
Input: transaction_amounts = [100, 110, 115, 90, 200, 85], n = 2
Output: [False, True]
Explanation: The model detects that there is a spike in the transaction after a series of normal transactions.Constraints:
- 1 <= transaction_amounts.length <= 10^5
- 0 <= transaction_amounts[i] <= 10^4