diff --git a/nn/include/feedforward.h b/nn/include/feedforward.h
new file mode 100644
index 0000000..fe76aec
--- /dev/null
+++ b/nn/include/feedforward.h
@@ -0,0 +1,14 @@
+#pragma once
+
+#include "neural_network.h"
+
+//==============================
+// Forward Pass Functions
+//==============================
+
+NeuralNetwork* create_network(int num_layers);
+
+void free_network(NeuralNetwork* nn);
+
+// It caches the intermediate results for the backpropagation algorithm.
+Matrix* feedforward(NeuralNetwork* nn, const Matrix* input);
diff --git a/nn/include/neural_network.h b/nn/include/neural_network.h
new file mode 100644
index 0000000..e735a93
--- /dev/null
+++ b/nn/include/neural_network.h
@@ -0,0 +1,33 @@
+#pragma once
+
+#include "activation.h"
+#include "cache.h"
+#include "linalg.h"
+
+//==============================
+// Neural Network Layer Struct
+//==============================
+
+typedef Matrix* (*ActivationFunc)(Matrix*);
+
+typedef struct {
+  Matrix* weights;
+  Matrix* bias;
+
+  ActivationFunc activation;
+
+  // The leak parameter for the Leaky ReLU activation function.
+  double leak_parameter;
+} Layer;
+
+//==============================
+// Neural Network Struct
+//==============================
+
+typedef struct {
+  Layer** layers;
+  int num_layers;
+
+  // A caching mechanism to store intermediate values from the forward pass.
+  Cache* cache;
+} NeuralNetwork;
diff --git a/nn/include/utils.h b/nn/include/utils.h
index 5252130..09572bf 100644
--- a/nn/include/utils.h
+++ b/nn/include/utils.h
@@ -29,6 +29,12 @@ typedef enum {
 // For safe malloc allocations
 #define CHECK_MALLOC(ptr, message) ASSERT((ptr) != NULL, message);
 
+// Macro to handle memory allocation checks.
+#define CHECK_ALLOC(ptr) \
+  if (ptr == NULL) {     \
+    return NULL;         \
+  }
+
 // Function prototypes for logging
 void log_message(LogLevel level, const char* format, ...);
 
diff --git a/nn/src/neural_network/feedforward.c b/nn/src/neural_network/feedforward.c
new file mode 100644
index 0000000..1e90e50
--- /dev/null
+++ b/nn/src/neural_network/feedforward.c
@@ -0,0 +1,96 @@
+#include "feedforward.h"
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "linalg.h"
+#include "neural_network.h"
+#include "utils.h"
+
+NeuralNetwork* create_network(int num_layers) {
+  NeuralNetwork* nn = (NeuralNetwork*)malloc(sizeof(NeuralNetwork));
+  if (nn == NULL) {
+    LOG_ERROR("Memory allocation failed for Neural Network struct.");
+    return NULL;
+  }
+
+  nn->layers = (Layer**)malloc(sizeof(Layer*) * num_layers);
+  if (nn->layers == NULL) {
+    LOG_ERROR("Memory allocation failed for layers array.");
+    free(nn);
+    return NULL;
+  }
+
+  nn->num_layers = num_layers;
+  nn->cache = init_cache();
+  if (nn->cache == NULL) {
+    LOG_ERROR("Failed to initialize cache.");
+    free(nn->layers);
+    free(nn);
+    return NULL;
+  }
+  return nn;
+}
+
+void free_network(NeuralNetwork* nn) {
+  if (nn == NULL) {
+    return;
+  }
+
+  if (nn->layers != NULL) {
+    for (int i = 0; i < nn->num_layers; i++) {
+      if (nn->layers[i] != NULL) {
+        if (nn->layers[i]->weights != NULL) {
+          free_matrix(nn->layers[i]->weights);
+        }
+        if (nn->layers[i]->bias != NULL) {
+          free_matrix(nn->layers[i]->bias);
+        }
+        free(nn->layers[i]);
+      }
+    }
+    free(nn->layers);
+  }
+  if (nn->cache != NULL) {
+    clear_cache(nn->cache);
+    free(nn->cache);
+  }
+  free(nn);
+}
+
+Matrix* feedforward(NeuralNetwork* nn, const Matrix* input) {
+  ASSERT(nn != NULL, "Neural Network pointer cannot be NULL.");
+  ASSERT(input != NULL, "Input matrix cannot be NULL.");
+  ASSERT(input->rows == nn->layers[0]->weights->cols,
+         "Input dimensions must match network dimensions.");
+
+  Matrix* current_output = copy_matrix(input);
+
+  // Cache the input for the backpropagation algorithm.
+  put_matrix(nn->cache, "input", current_output);
+
+  for (int i = 0; i < nn->num_layers; i++) {
+    Matrix* z = dot_matrix(current_output, nn->layers[i]->weights);
+
+    add_matrix(z, nn->layers[i]->bias);
+
+    // Cache the intermediate value (z).
+    char z_key[32];
+    sprintf(z_key, "z_%d", i);
+    put_matrix(nn->cache, z_key, z);
+
+    Matrix* a = apply_onto_matrix(nn->layers[i]->activation, z);
+
+    // Cache the activated output (a).
+    char a_key[32];
+    sprintf(a_key, "a_%d", i);
+    put_matrix(nn->cache, a_key, a);
+
+    free_matrix(z);
+    free_matrix(current_output);
+    current_output = a;
+  }
+
+  return current_output;
+}