How NumPy Helps You Land a Data Scientist Job Faster

import numpy as np

spend = np.array([0, 120, 4500, 90, 250000, 640], dtype=float)
visits = np.array([0, 3, 10, 2, 12, 4], dtype=float)

log_spend = np.log1p(spend)
capped = np.clip(spend, 0, np.percentile(spend, 95))
per_visit = np.divide(spend, visits, out=np.zeros_like(spend), where=visits != 0)
print(log_spend.round(2), capped.round(0), per_visit.round(1), sep="\n")

import numpy as np

ages = np.array([17, 24, 35, 52, 61, 29])
is_senior = ages >= 60              # boolean mask
bucket = np.where(ages < 25, "young", np.where(ages < 50, "mid", "senior"))
print(is_senior)
print(bucket)
print("share senior:", is_senior.mean())

import numpy as np

rng = np.random.default_rng(42)
X = rng.normal(size=(200, 2))
true_w = np.array([3.0, -1.5])
y = X @ true_w + 4.0 + rng.normal(scale=0.5, size=200)

X_b = np.column_stack([np.ones(len(X)), X])   # add intercept column
w, *_ = np.linalg.lstsq(X_b, y, rcond=None)
print("intercept, w1, w2:", w.round(2))       # close to 4, 3, -1.5

import numpy as np

points = np.array([[1, 2], [3, 4], [10, 10]], dtype=float)
query = np.array([2, 3], dtype=float)
dists = np.linalg.norm(points - query, axis=1)   # broadcasting
nearest = np.argmin(dists)
print(dists.round(2), "nearest index:", nearest)

import numpy as np

rng = np.random.default_rng(7)
orders = np.array([450, 820, 390, 1200, 610, 700, 980, 530, 875, 660])
boot_means = np.array([
    rng.choice(orders, size=orders.size, replace=True).mean()
    for _ in range(5000)
])
low, high = np.percentile(boot_means, [2.5, 97.5])
print(f"mean {orders.mean():.0f}, 95% CI {low:.0f} to {high:.0f}")

import numpy as np

rng = np.random.default_rng(1)
a = np.array([12, 15, 11, 14, 13, 16])     # variant A metric
b = np.array([15, 17, 16, 18, 14, 19])     # variant B metric
observed = b.mean() - a.mean()
pooled = np.concatenate([a, b])
count = 0
for _ in range(10000):
    p = rng.permutation(pooled)
    if p[len(a):].mean() - p[:len(a)].mean() >= observed:
        count += 1
print("observed diff:", observed.round(2), "p-value approx:", count / 10000)

NumPy exercises
1. Normalise each column of a 2D array to mean 0, std 1
2. Find the indexes of the 3 largest values
3. Replace negative values with 0 without a loop
4. Compute pairwise distances between rows of a small matrix
5. Bootstrap a 90% interval for a median

import numpy as np
X = np.random.default_rng(0).normal(size=(5, 3))
Z = (X - X.mean(axis=0)) / X.std(axis=0)            # 1
top3 = np.argsort(X.ravel())[-3:][::-1]              # 2
Xpos = np.where(X < 0, 0, X)                         # 3
D = np.linalg.norm(X[:, None, :] - X[None, :, :], axis=2)   # 4
print(Z.round(2), top3, Xpos.round(2), D.shape, sep="\n")

Shape checklist (say it aloud in interviews)
- What is the shape of each input?
- Which axis am I reducing over?
- Will broadcasting align from the right?
- What shape should the output have?
- Print .shape once to confirm