Not a course. A conversation that teaches.

import numpy as np

class NeuralNet:
    def __init__(self, dims):
        self.W = [np.random.randn(d1,d2)*0.01
                  for d1,d2 in zip(dims,dims[1:])]
        self.b = [np.zeros((1,d)) for d in dims[1:]]

    def forward(self, X):
        self.a = [X]
        for W,b in zip(self.W, self.b):
            z = self.a[-1] @ W + b
            self.a.append(np.maximum(0, z))  # ReLU
        return self.a[-1]

    def backward(self, y_true, lr=0.01):
        m = len(y_true)
        dout = self.a[-1] - y_true
        for i in reversed(range(len(self.W))):
            dw = self.a[i].T @ dout / m
            db = np.sum(dout, axis=0, keepdims=True) / m
            if i > 0:
                dout = (dout @ self.W[i].T)
                dout *= (self.a[i] > 0)
            self.W[i] -= lr * dw
            self.b[i] -= lr * db

    def train(self, X, y, epochs=1000, lr=0.01):
        for _ in range(epochs):
            self.forward(X)
            self.backward(y, lr)