Meta Data Scientist Interview Questions

Meta requires a robust way to manage various versions of machine learning models used across different services. The Model Versioning Registry should provide capabilities to track changes, roll back, and retrieve model metadata efficiently.
Problem statement: Implement a class ModelVersioningRegistry that handles the versioning of ML models:
- def __init__(self): Constructs an empty model registry.
- def add_model(self, model_id: str, version: str, metadata: Dict[str, Any]) -> None: Adds a new model version to the registry.
- def get_model(self, model_id: str, version: str) -> Dict[str, Any]: Retrieves metadata for a specific model version from the registry.
- def rollback_model(self, model_id: str) -> str: Rolls back to the previous version of the specified model.
Constraints:
- 1 <= len(model_id) <= 255
- 1 <= len(version) <= 50

Meta wants to analyze and simulate the growth patterns of user engagement on Instagram Reels as different features go live. Design a simulator to project user adoption and interactions over a given time period based on historical usage data.
Problem statement: Create a class FeatureGrowthSimulator that simulates viral growth patterns for Instagram features:
- def __init__(self, initial_users: int, growth_rate: float): Constructs the simulator with initial number of users and growth rate.
- def simulate_growth(self, months: int) -> List[int]: Simulates the growth for the given number of months and returns a list of user counts at the end of each month.
- def set_growth_rate(self, new_growth_rate: float) -> None: Updates the growth rate based on external factors.
Constraints:
- 1 <= initial_users <= 10^6
- 0 < growth_rate <= 1
- 1 <= months <= 24

Meta aims to improve communication speed on Messenger by analyzing network distances. You need to determine the shortest path between users based on a network graph where nodes represent users and edges represent direct communication paths with weights representing latency.
Problem statement: Implement Dijkstra's algorithm to find the shortest path from a given starting node to all other nodes in a weighted graph.
- Method Signature: def dijkstra(graph: Dict[int, List[Tuple[int, int]]], start: int) -> Dict[int, int]
Example 1:
Input: graph = {0: [(1, 1), (2, 4)], 1: [(2, 2)], 2: []}, start = 0
Output: {0: 0, 1: 1, 2: 3}
Explanation: The shortest path from node 0 to node 1 is 1 and to node 2 is 3 (0 -> 1 -> 2).
Constraints:
- 1 <= len(graph) <= 1000
- 0 <= weights <= 1000

Meta needs to optimize user engagement on Facebook Live. You are tasked with analyzing user interaction data in real-time to identify active segments of the audience during a live event.
Problem statement: Given an array of integers engagements where each integer represents the number of interactions at each timestamp, determine the maximum number of interactions in any k consecutive timestamps.
- Method Signature: def max_engagement(engagements: List[int], k: int) -> int
Example 1:
Input: engagements = [1, 2, 3, 4, 5], k = 2
Output: 9
Explanation: The maximum interactions in any 2 consecutive timestamps is 4 + 5 = 9.
Constraints:
- 1 <= len(engagements) <= 10^5
- 1 <= k <= len(engagements)