NumPy Interview Questions

This document provides a curated list of NumPy interview questions commonly asked in technical interviews for Data Science, Quantitative Analyst, Machine Learning Engineer, and High-Performance Computing roles. It covers everything from array manipulation to advanced linear algebra and memory management.

This is updated frequently but right now this is the most exhaustive list of type of questions being asked.

Code Examples

1. Advanced Broadcasting

import numpy as np

# Calculating distance matrix between two sets of points
# A: (3, 2), B: (4, 2)
A = np.array([[1,1], [2,2], [3,3]])
B = np.array([[4,4], [5,5], [6,6], [7,7]])

# Shape manipulation for broadcasting
# shape (3,1,2) - shape (1,4,2) -> shape (3,4,2)
diff = A[:, np.newaxis, :] - B[np.newaxis, :, :]

# Summing squares along last axis: shape (3,4)
dists = np.sum(diff**2, axis=-1)
print(dists)

2. Efficient Sliding Window (Stride Tricks)

import numpy as np
from numpy.lib.stride_tricks import as_strided

def sliding_window(arr, window_size):
    """
    Efficiently create sliding windows without copying data.
    """
    stride = arr.strides[0]
    shape = (len(arr) - window_size + 1, window_size)
    strides = (stride, stride)
    return as_strided(arr, shape=shape, strides=strides)

arr = np.arange(10)
print(sliding_window(arr, 3))

3. Einstein Summation

import numpy as np

A = np.random.rand(2, 3)
B = np.random.rand(3, 4)
C = np.random.rand(2, 4)

# Matrix multiplication: A @ B
res_matmul = np.einsum('ik,kj->ij', A, B)

# Dot product of rows in A and C
res_dot = np.einsum('ij,ij->i', A, C)

print("Matmul shape:", res_matmul.shape)
print("Row dot shape:", res_dot.shape)

Questions asked in Google interview

• How would you implement convolution from scratch using stride tricks?
• Explain the memory layout of C vs Fortran arrays in NumPy
• Write code to efficiently calculate pairwise distances
• How to handle numerical stability in large matrix operations?
• Explain broadcasting rules with examples
• How would you optimize a slow loop over arrays?
• Write code to perform image padding manually
• How to implement moving average without loops?
• Explain the usage of einsum vs dot product
• How to handle large datasets that don't fit in RAM?

Questions asked in Amazon interview

• Write code to implement sparse matrix multiplication
• How would you generate non-uniform random numbers?
• Explain vectorized boolean operations
• Write code to filter values without creating a copy
• How to optimize array concatenation in a loop?
• Explain eigen decomposition implementation details
• Write code to solve system of linear equations
• How to handle missing values in numeric arrays?
• Explain performance difference between float32 vs float64
• Write code to normalize a matrix row-wise

Questions asked in Meta interview

• How would you implement efficient array sorting?
• Explain structured arrays and their use cases
• Write code to compute histograms on multidimensional data
• How to implement custom reduction functions?
• Explain caching effects on array operations
• Write code to rotate an image represented as an array
• How to handle overflow in integer arrays?
• Explain how ufuncs work internally
• Write code to efficiently slicing multi-dimensional arrays
• How to implement vectorized string operations?

Questions asked in Microsoft interview

• Explain the role of BLAS/LAPACK in NumPy
• Write code to compute the inverse of a matrix
• How to create a view of an array with different data type?
• Explain memory mapping for large files
• Write code to perform fast fourier transform
• How to implement a custom random number generator?
• Explain broadcasting errors and how to fix them
• Write code to compute cross-correlation
• How to optimize dot product for sparse vectors?
• Explain how to use np.where for complex conditions

Questions asked in HFT Firms (e.g., Jane Street, Citadel)

• How to optimize stride usage for cache locality?
• Write code to implement order management system logic with arrays
• Explain floating point precision pitfalls in financial calc
• How to minimize memory allocations in critical paths?
• Write code to implement rolling window statistics efficiently
• how to use Numba to accelerate NumPy logic?
• Explain SIMD instructions usage in NumPy
• Write code to process tick data efficiently
• How to handle NaN propagation in accumulation?
• Explain the difference between np.random.rand and np.random.Generator

Additional Resources

• Official NumPy Documentation
• From Python to NumPy
• 100 NumPy Exercises
• Scipy Lecture Notes
• NumPy Visualization