diff --git a/examples/codegen/README.md b/examples/codegen/README.md
new file mode 100644
index 0000000000..e28b9aa41f
--- /dev/null
+++ b/examples/codegen/README.md
@@ -0,0 +1,35 @@
+## End-to-end code generation example
+
+This example demonstrates an end-to-end pipeline of code generation and execution that typically occurs in a JIT engine.
+It creates a simple function (adding +1 to an array and storing the result in a second one) via llvm API,
+converts it to spirv, and submits it to the runtime.
+
+## Running the example
+
+The following commands are executed from the project root directory.
+
+### Using conda environment
+
+```
+$ conda env create -f third_party/deps.yml
+$ conda activate examples
+$ mkdir build && cd build
+# configure with:
+$ cmake .. -DUR_BUILD_ADAPTER_L0=ON -DUR_BUILD_EXAMPLE_CODEGEN=ON
+$ make
+$ ./bin/codegen
+```
+
+### Without using conda
+
+To run the example with llvm 13, you will need to make sure that `llvm-13` and `libllvmspirvlib-13-dev` are installed.  
+
+**NOTE**: The example could be working with other llvm versions but it was tested with version 13.
+
+```
+$ mkdir build && cd build
+# configure with:
+$ cmake .. -DUR_BUILD_ADAPTER_L0=ON -DUR_BUILD_EXAMPLE_CODEGEN=ON -DLLVM_DIR=/usr/lib/llvm-13/cmake
+$ make
+$ ./bin/codegen
+```
diff --git a/examples/codegen/codegen.cpp b/examples/codegen/codegen.cpp
index 0352ae6e02..179c8c923e 100644
--- a/examples/codegen/codegen.cpp
+++ b/examples/codegen/codegen.cpp
@@ -20,6 +20,8 @@
 #include "helpers.h"
 #include "ur_api.h"
 
+const unsigned PAGE_SIZE = 4096;
+
 void ur_check(const ur_result_t r) {
     if (r != UR_RESULT_SUCCESS) {
         urTearDown(nullptr);
@@ -40,6 +42,23 @@ std::vector<ur_adapter_handle_t> get_adapters() {
     return adapters;
 }
 
+std::vector<ur_adapter_handle_t>
+get_supported_adapters(std::vector<ur_adapter_handle_t> &adapters) {
+    std::vector<ur_adapter_handle_t> supported_adapters;
+    for (auto adapter : adapters) {
+        ur_adapter_backend_t backend;
+        ur_check(urAdapterGetInfo(adapter, UR_ADAPTER_INFO_BACKEND,
+                                  sizeof(ur_adapter_backend_t), &backend,
+                                  nullptr));
+
+        if (backend == UR_ADAPTER_BACKEND_LEVEL_ZERO) {
+            supported_adapters.push_back(adapter);
+        }
+    }
+
+    return supported_adapters;
+}
+
 std::vector<ur_platform_handle_t>
 get_platforms(std::vector<ur_adapter_handle_t> &adapters) {
     uint32_t platformCount = 0;
@@ -70,7 +89,7 @@ std::vector<ur_device_handle_t> get_gpus(ur_platform_handle_t p) {
     return devices;
 }
 
-template <typename T, size_t N> struct alignas(4096) AlignedArray {
+template <typename T, size_t N> struct alignas(PAGE_SIZE) AlignedArray {
     T data[N];
 };
 
@@ -79,7 +98,8 @@ int main() {
     ur_check(urInit(UR_DEVICE_INIT_FLAG_GPU, loader_config));
 
     auto adapters = get_adapters();
-    auto platforms = get_platforms(adapters);
+    auto supported_adapters = get_supported_adapters(adapters);
+    auto platforms = get_platforms(supported_adapters);
     auto gpus = get_gpus(platforms.front());
     auto spv = generate_plus_one_spv();
 
@@ -131,10 +151,26 @@ int main() {
 
     ur_check(urQueueFinish(queue));
 
+    std::cout << "Input Array: ";
+    for (int i = 0; i < a_size; ++i) {
+        std::cout << a.data[i] << " ";
+    }
+    std::cout << std::endl;
+
+    bool expectedResult = false;
+
+    std::cout << "Output Array: ";
     for (int i = 0; i < a_size; ++i) {
         std::cout << b.data[i] << " ";
+        expectedResult |= (b.data[i] == a.data[i] + 1);
     }
     std::cout << std::endl;
 
-    return urTearDown(nullptr) == UR_RESULT_SUCCESS ? 0 : 1;
+    if (expectedResult) {
+        std::cout << "Results are correct." << std::endl;
+    } else {
+        std::cout << "Results are incorrect." << std::endl;
+    }
+
+    return urTearDown(nullptr) == UR_RESULT_SUCCESS && expectedResult ? 0 : 1;
 }
diff --git a/scripts/deps.yml b/third_party/deps.yml
similarity index 100%
rename from scripts/deps.yml
rename to third_party/deps.yml