Why Every Python Developer Should Learn NumPy in 2026

import numpy as np

prices = [120, 250, 80, 400]
arr = np.array(prices)

print([p * 1.18 for p in prices])   # list: needs a loop
print(arr * 1.18)                    # array: one expression
print(arr.dtype, arr.shape, arr.mean(), arr.max())

matrix = np.array([[1, 2, 3], [4, 5, 6]])
print(matrix.shape)          # (2, 3)
print(matrix.T)              # transpose
print(matrix.sum(axis=0))    # column sums

import time
import numpy as np

n = 1_000_000
lst = list(range(n))
arr = np.arange(n)

t = time.perf_counter(); s1 = sum(x * x for x in lst); t_list = time.perf_counter() - t
t = time.perf_counter(); s2 = int((arr * arr).sum()); t_np = time.perf_counter() - t

print("same result:", s1 == s2)
print(f"list: {t_list:.3f}s  numpy: {t_np:.4f}s")
# Exact timings depend on your machine.

import numpy as np

orders = np.array([450, 1200, 80, 2600, 990, 150])

big = orders[orders > 1000]              # boolean mask
print("big orders:", big)
print("share of big orders:", (orders > 1000).mean())

discounted = np.where(orders > 1000, orders * 0.9, orders)
print("after discount:", discounted)
print("total, mean, median:", orders.sum(), orders.mean().round(1), np.median(orders))

import numpy as np

# rows = students, columns = subjects
scores = np.array([[70, 85, 90],
                   [60, 75, 80],
                   [90, 95, 85]], dtype=float)

subject_mean = scores.mean(axis=0)           # shape (3,)
centred = scores - subject_mean              # (3,3) - (3,) broadcasts
z = centred / scores.std(axis=0)             # standardise per subject
print(subject_mean)
print(z.round(2))

import numpy as np
import pandas as pd

df = pd.DataFrame({"amount": [500, 750, 300], "qty": [1, 3, 2]})
arr = df.to_numpy()
print(type(arr), arr.dtype, arr.shape)

# Vectorised NumPy maths on DataFrame columns
df["unit_price"] = np.round(df["amount"] / df["qty"], 2)
df["log_amount"] = np.log1p(df["amount"])
print(df)

import numpy as np
from sklearn.linear_model import LinearRegression

x = np.array([1, 2, 3, 4, 5])          # shape (5,)
y = np.array([3.1, 5.0, 7.2, 8.9, 11.1])

# Models need 2D input: (rows, features)
X = x.reshape(-1, 1)                    # shape (5, 1)
model = LinearRegression().fit(X, y)
print("slope:", model.coef_.round(2), "intercept:", round(model.intercept_, 2))
print("prediction for 6:", model.predict(np.array([[6]])).round(2))

1. Creating arrays: np.array, arange, linspace, zeros, ones
2. dtype, shape, ndim, reshape
3. Indexing and slicing (1D and 2D)
4. Boolean masks and np.where
5. Aggregations with axis
6. Broadcasting rules
7. Random numbers with np.random.default_rng
8. Saving and loading: np.save, np.load, np.loadtxt

Days 1-2   Creating arrays, dtype, shape, reshape
Days 3-4   Indexing, slicing, masks
Days 5-6   Aggregations and axis
Days 7-8   Broadcasting exercises
Days 9-10  Random data and simulations
Days 11-12 Use NumPy with pandas and scikit-learn
Days 13-14 Mini project: vectorise a slow loop-based script

1. Create 1,000 random exam scores (0-100); find mean, median, top 10%
2. Convert a list of rupee prices to dollars with one expression
3. From a 2D array of daily sales (stores x days), find each store's best day
4. Standardise each column of a 2D array (mean 0, std 1)
5. Replace all negative values in an array with 0 without a loop
6. Rewrite a nested-loop distance calculation using broadcasting