diff --git a/CMakeLists.txt b/CMakeLists.txt
index 0ff56f890b4e01547de322c0aed45e4887bae344..822c1137e7398f7dd71caac0797f1dbf0f900e22 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -57,25 +57,25 @@ set(KLEIDIAI_WARNING_FLAGS
 )
 
 set(KLEIDIAI_FILES_NEON
-    src/matmul/kai_lhs_quant_pack_qai8dxp_f32.c
-    src/matmul/kai_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0.c
+    kai/ukernels/matmul/pack/kai_lhs_quant_pack_qai8dxp_f32.c
+    kai/ukernels/matmul/pack/kai_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0.c
 )
 
 set(KLEIDIAI_FILES_NEON_FP16
-    src/matmul/kai_rhs_pack_kxn_f16p16x1biasf16_f16_f16_neon.c
-    src/matmul/matmul_clamp_f16_f16_f16p/kai_matmul_clamp_f16_f16_f16p16x1biasf16_6x16x8_neon_mla.c
+    kai/ukernels/matmul/pack/kai_rhs_pack_kxn_f16p16x1biasf16_f16_f16_neon.c
+    kai/ukernels/matmul/matmul_clamp_f16_f16_f16p/kai_matmul_clamp_f16_f16_f16p16x1biasf16_6x16x8_neon_mla.c
 )
 
 set(KLEIDIAI_FILES_NEON_DOTPROD
-    src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp4x8_1x4x32_neon_dotprod.c
-    src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp8x8_1x8x32_neon_dotprod.c
+    kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp4x8_1x4x32_neon_dotprod.c
+    kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp8x8_1x8x32_neon_dotprod.c
 )
 
 set(KLEIDIAI_FILES_NEON_I8MM
-    src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_4x4x32_neon_i8mm.c
-    src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_8x4x32_neon_i8mm.c
-    src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_4x8x32_neon_i8mm.c
-    src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_8x8x32_neon_i8mm.c
+    kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_4x4x32_neon_i8mm.c
+    kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_8x4x32_neon_i8mm.c
+    kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_4x8x32_neon_i8mm.c
+    kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_8x8x32_neon_i8mm.c
 )
 
 add_library(kleidiai
@@ -86,8 +86,7 @@ add_library(kleidiai
 )
 
 target_include_directories(kleidiai
-    PUBLIC  src
-    PRIVATE .
+    PUBLIC  .
 )
 
 target_compile_options(kleidiai
@@ -142,10 +141,6 @@ if(KLEIDIAI_BUILD_TESTS)
         test/tests/matmul_test.cpp
     )
 
-    target_include_directories(kleidiai_test
-        PRIVATE .
-    )
-
     target_compile_options(kleidiai_test
         PRIVATE ${KLEIDIAI_WARNING_FLAGS}
         PRIVATE -march=armv8.2-a+fp16+bf16
diff --git a/src/kai_common.h b/kai/kai_common.h
similarity index 100%
rename from src/kai_common.h
rename to kai/kai_common.h
diff --git a/src/matmul/matmul_clamp_f16_f16_f16p/kai_matmul_clamp_f16_f16_f16p16x1biasf16_6x16x8_neon_mla.c b/kai/ukernels/matmul/matmul_clamp_f16_f16_f16p/kai_matmul_clamp_f16_f16_f16p16x1biasf16_6x16x8_neon_mla.c
similarity index 99%
rename from src/matmul/matmul_clamp_f16_f16_f16p/kai_matmul_clamp_f16_f16_f16p16x1biasf16_6x16x8_neon_mla.c
rename to kai/ukernels/matmul/matmul_clamp_f16_f16_f16p/kai_matmul_clamp_f16_f16_f16p16x1biasf16_6x16x8_neon_mla.c
index c4996ebd55addce61d09cc9a33e9713f80b48444..8c848afd1b602d64d4859da4987e95ca0006a6e7 100644
--- a/src/matmul/matmul_clamp_f16_f16_f16p/kai_matmul_clamp_f16_f16_f16p16x1biasf16_6x16x8_neon_mla.c
+++ b/kai/ukernels/matmul/matmul_clamp_f16_f16_f16p/kai_matmul_clamp_f16_f16_f16p16x1biasf16_6x16x8_neon_mla.c
@@ -13,7 +13,7 @@
 #include <stddef.h>
 #include <stdint.h>
 
-#include "kai_common.h"
+#include "kai/kai_common.h"
 
 static const size_t kai_mr = 6;
 static const size_t kai_nr = 16;
diff --git a/src/matmul/matmul_clamp_f16_f16_f16p/kai_matmul_clamp_f16_f16_f16p16x1biasf16_6x16x8_neon_mla.h b/kai/ukernels/matmul/matmul_clamp_f16_f16_f16p/kai_matmul_clamp_f16_f16_f16p16x1biasf16_6x16x8_neon_mla.h
similarity index 100%
rename from src/matmul/matmul_clamp_f16_f16_f16p/kai_matmul_clamp_f16_f16_f16p16x1biasf16_6x16x8_neon_mla.h
rename to kai/ukernels/matmul/matmul_clamp_f16_f16_f16p/kai_matmul_clamp_f16_f16_f16p16x1biasf16_6x16x8_neon_mla.h
diff --git a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp4x8_1x4x32_neon_dotprod.c b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp4x8_1x4x32_neon_dotprod.c
similarity index 99%
rename from src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp4x8_1x4x32_neon_dotprod.c
rename to kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp4x8_1x4x32_neon_dotprod.c
index 4d47ef1e947dd947d05aa31741dd67412d27ec3f..fa9a3f5c22b42dca283d5fa154e76890d7f40b17 100644
--- a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp4x8_1x4x32_neon_dotprod.c
+++ b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp4x8_1x4x32_neon_dotprod.c
@@ -11,7 +11,7 @@
 #include <arm_neon.h>
 #include <stdint.h>
 
-#include "kai_common.h"
+#include "kai/kai_common.h"
 
 static const size_t kai_m_step = 1;
 static const size_t kai_n_step = 4;
diff --git a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp4x8_1x4x32_neon_dotprod.h b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp4x8_1x4x32_neon_dotprod.h
similarity index 100%
rename from src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp4x8_1x4x32_neon_dotprod.h
rename to kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp4x8_1x4x32_neon_dotprod.h
diff --git a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp8x8_1x8x32_neon_dotprod.c b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp8x8_1x8x32_neon_dotprod.c
similarity index 99%
rename from src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp8x8_1x8x32_neon_dotprod.c
rename to kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp8x8_1x8x32_neon_dotprod.c
index 6bff095a3d2babb95df7923f3fd1de076c67ca3d..17707488d294b27775755f199f94c93fc8d1b5e5 100644
--- a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp8x8_1x8x32_neon_dotprod.c
+++ b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp8x8_1x8x32_neon_dotprod.c
@@ -11,7 +11,7 @@
 #include <arm_neon.h>
 #include <stdint.h>
 
-#include "kai_common.h"
+#include "kai/kai_common.h"
 
 static const size_t kai_m_step = 1;
 static const size_t kai_n_step = 8;
diff --git a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp8x8_1x8x32_neon_dotprod.h b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp8x8_1x8x32_neon_dotprod.h
similarity index 100%
rename from src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp8x8_1x8x32_neon_dotprod.h
rename to kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp1x8_qsi4cxp8x8_1x8x32_neon_dotprod.h
diff --git a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_4x4x32_neon_i8mm.c b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_4x4x32_neon_i8mm.c
similarity index 99%
rename from src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_4x4x32_neon_i8mm.c
rename to kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_4x4x32_neon_i8mm.c
index 45f431707eff835a24304da83a01e4422ff80609..ebcaaf544db70b42bd4765adc46737908a2b40b9 100644
--- a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_4x4x32_neon_i8mm.c
+++ b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_4x4x32_neon_i8mm.c
@@ -11,7 +11,7 @@
 #include <arm_neon.h>
 #include <stdint.h>
 
-#include "kai_common.h"
+#include "kai/kai_common.h"
 
 static const size_t kai_m_step = 4;
 static const size_t kai_n_step = 4;
diff --git a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_4x4x32_neon_i8mm.h b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_4x4x32_neon_i8mm.h
similarity index 100%
rename from src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_4x4x32_neon_i8mm.h
rename to kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_4x4x32_neon_i8mm.h
diff --git a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_8x4x32_neon_i8mm.c b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_8x4x32_neon_i8mm.c
similarity index 99%
rename from src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_8x4x32_neon_i8mm.c
rename to kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_8x4x32_neon_i8mm.c
index 293fad2907affa2d79e7e68e9a4c165bb05d00de..ac9ec40e064b9080d7c7901f0d40381132c48bec 100644
--- a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_8x4x32_neon_i8mm.c
+++ b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_8x4x32_neon_i8mm.c
@@ -11,7 +11,7 @@
 #include <arm_neon.h>
 #include <stdint.h>
 
-#include "kai_common.h"
+#include "kai/kai_common.h"
 
 static const size_t kai_m_step = 8;
 static const size_t kai_n_step = 4;
diff --git a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_8x4x32_neon_i8mm.h b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_8x4x32_neon_i8mm.h
similarity index 100%
rename from src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_8x4x32_neon_i8mm.h
rename to kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp4x8_8x4x32_neon_i8mm.h
diff --git a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_4x8x32_neon_i8mm.c b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_4x8x32_neon_i8mm.c
similarity index 99%
rename from src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_4x8x32_neon_i8mm.c
rename to kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_4x8x32_neon_i8mm.c
index 06adc07a8ae7d8d37ff64bf14915d2d713cf7020..93c1a350a34f5416d0a4dcab4aa27dd2df3e2a96 100644
--- a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_4x8x32_neon_i8mm.c
+++ b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_4x8x32_neon_i8mm.c
@@ -11,7 +11,7 @@
 #include <arm_neon.h>
 #include <stdint.h>
 
-#include "kai_common.h"
+#include "kai/kai_common.h"
 
 static const size_t kai_m_step = 4;
 static const size_t kai_n_step = 8;
diff --git a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_4x8x32_neon_i8mm.h b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_4x8x32_neon_i8mm.h
similarity index 100%
rename from src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_4x8x32_neon_i8mm.h
rename to kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_4x8x32_neon_i8mm.h
diff --git a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_8x8x32_neon_i8mm.c b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_8x8x32_neon_i8mm.c
similarity index 99%
rename from src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_8x8x32_neon_i8mm.c
rename to kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_8x8x32_neon_i8mm.c
index 9d8153a98b0608584d433ccc2d403898d4d33282..2b2b1e09b7c92adebe07cabe9a1217f87217d2cd 100644
--- a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_8x8x32_neon_i8mm.c
+++ b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_8x8x32_neon_i8mm.c
@@ -11,7 +11,7 @@
 #include <arm_neon.h>
 #include <stdint.h>
 
-#include "kai_common.h"
+#include "kai/kai_common.h"
 
 static const size_t kai_m_step = 8;
 static const size_t kai_n_step = 8;
diff --git a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_8x8x32_neon_i8mm.h b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_8x8x32_neon_i8mm.h
similarity index 100%
rename from src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_8x8x32_neon_i8mm.h
rename to kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_8x8x32_neon_i8mm.h
diff --git a/src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp_qsi4cxp_interface.h b/kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp_qsi4cxp_interface.h
similarity index 100%
rename from src/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp_qsi4cxp_interface.h
rename to kai/ukernels/matmul/matmul_clamp_f32_qai8dxp_qsi4cxp/kai_matmul_clamp_f32_qai8dxp_qsi4cxp_interface.h
diff --git a/src/matmul/kai_lhs_quant_pack_qai8dxp_f32.c b/kai/ukernels/matmul/pack/kai_lhs_quant_pack_qai8dxp_f32.c
similarity index 99%
rename from src/matmul/kai_lhs_quant_pack_qai8dxp_f32.c
rename to kai/ukernels/matmul/pack/kai_lhs_quant_pack_qai8dxp_f32.c
index 6e618ccc08f130824ce0cbc31d51f97fc12487ac..ffb09949ae227b4bdbb5cc834357086eecf0d90e 100644
--- a/src/matmul/kai_lhs_quant_pack_qai8dxp_f32.c
+++ b/kai/ukernels/matmul/pack/kai_lhs_quant_pack_qai8dxp_f32.c
@@ -10,7 +10,7 @@
 #include <math.h>
 #include <stdint.h>
 
-#include "kai_common.h"
+#include "kai/kai_common.h"
 
 static const size_t kai_num_bytes_per_multiplier = sizeof(float);
 static const size_t kai_num_bytes_per_offset = sizeof(int32_t);
diff --git a/src/matmul/kai_lhs_quant_pack_qai8dxp_f32.h b/kai/ukernels/matmul/pack/kai_lhs_quant_pack_qai8dxp_f32.h
similarity index 100%
rename from src/matmul/kai_lhs_quant_pack_qai8dxp_f32.h
rename to kai/ukernels/matmul/pack/kai_lhs_quant_pack_qai8dxp_f32.h
diff --git a/src/matmul/kai_rhs_pack_kxn_f16p16x1biasf16_f16_f16_neon.c b/kai/ukernels/matmul/pack/kai_rhs_pack_kxn_f16p16x1biasf16_f16_f16_neon.c
similarity index 99%
rename from src/matmul/kai_rhs_pack_kxn_f16p16x1biasf16_f16_f16_neon.c
rename to kai/ukernels/matmul/pack/kai_rhs_pack_kxn_f16p16x1biasf16_f16_f16_neon.c
index 4bf043039295fc0cf150acf732463a681a53e69b..50c04ad3fe5ecb86a9bcc523f0dccbcd0ba445c3 100644
--- a/src/matmul/kai_rhs_pack_kxn_f16p16x1biasf16_f16_f16_neon.c
+++ b/kai/ukernels/matmul/pack/kai_rhs_pack_kxn_f16p16x1biasf16_f16_f16_neon.c
@@ -11,7 +11,7 @@
 #include <stddef.h>
 #include <stdint.h>
 
-#include "kai_common.h"
+#include "kai/kai_common.h"
 
 static const size_t kai_nr = 16;
 static const size_t kai_kr = 1;
diff --git a/src/matmul/kai_rhs_pack_kxn_f16p16x1biasf16_f16_f16_neon.h b/kai/ukernels/matmul/pack/kai_rhs_pack_kxn_f16p16x1biasf16_f16_f16_neon.h
similarity index 100%
rename from src/matmul/kai_rhs_pack_kxn_f16p16x1biasf16_f16_f16_neon.h
rename to kai/ukernels/matmul/pack/kai_rhs_pack_kxn_f16p16x1biasf16_f16_f16_neon.h
diff --git a/src/matmul/kai_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0.c b/kai/ukernels/matmul/pack/kai_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0.c
similarity index 99%
rename from src/matmul/kai_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0.c
rename to kai/ukernels/matmul/pack/kai_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0.c
index aed8a5f427b34f09a90805a0c8b4c62245a3da15..41d3fcdbc0331fc5637c8df050211d7c5bf9cb0e 100644
--- a/src/matmul/kai_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0.c
+++ b/kai/ukernels/matmul/pack/kai_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0.c
@@ -10,7 +10,7 @@
 #include <stdint.h>
 #include <string.h>
 
-#include "kai_common.h"
+#include "kai/kai_common.h"
 
 static const size_t kai_num_bytes_sum_rhs = sizeof(int32_t);
 static const size_t kai_num_bytes_multiplier_rhs = sizeof(float);
diff --git a/src/matmul/kai_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0.h b/kai/ukernels/matmul/pack/kai_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0.h
similarity index 100%
rename from src/matmul/kai_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0.h
rename to kai/ukernels/matmul/pack/kai_rhs_pack_nxk_qsi4cxp_qsu4cxs1s0.h
diff --git a/test/common/compare.cpp b/test/common/compare.cpp
index 9dcfb104bc2d4e0409bd9b4649585174fb11f9f1..91b6bc6656c6a6a95bbb8e465ab22760e4e3884f 100644
--- a/test/common/compare.cpp
+++ b/test/common/compare.cpp
@@ -12,7 +12,7 @@
 #include <tuple>
 #include <type_traits>
 
-#include "src/kai_common.h"
+#include "kai/kai_common.h"
 #include "test/common/bfloat16.hpp"
 #include "test/common/data_format.hpp"
 #include "test/common/data_type.hpp"
diff --git a/test/common/data_format.cpp b/test/common/data_format.cpp
index a69a085dc0b438bfd2b56ced081c02797904f92d..a16269e9dea6646a5f95852426ea672983dff9ab 100644
--- a/test/common/data_format.cpp
+++ b/test/common/data_format.cpp
@@ -9,7 +9,7 @@
 #include <cstddef>
 #include <cstdint>
 
-#include "src/kai_common.h"
+#include "kai/kai_common.h"
 #include "test/common/data_type.hpp"
 #include "test/reference/round.hpp"
 
diff --git a/test/common/data_type.cpp b/test/common/data_type.cpp
index 952097502e4677bff270e746273f016195428e2d..44f594afec4375bac91dc590e2d6bfc52019b091 100644
--- a/test/common/data_type.cpp
+++ b/test/common/data_type.cpp
@@ -9,7 +9,7 @@
 #include <cstddef>
 #include <cstdint>
 
-#include "src/kai_common.h"
+#include "kai/kai_common.h"
 
 namespace kai::test {
 
diff --git a/test/common/int4.cpp b/test/common/int4.cpp
index ae0a5998259879622906a422a6c32d889adc4d16..5f8afef59f3b0c943020674da8036b1b12a9fb6e 100644
--- a/test/common/int4.cpp
+++ b/test/common/int4.cpp
@@ -9,7 +9,7 @@
 #include <cstdint>
 #include <tuple>
 
-#include "kai_common.h"
+#include "kai/kai_common.h"
 
 namespace kai::test {
 
diff --git a/test/common/matrix_portion.cpp b/test/common/matrix_portion.cpp
index e7c42bf130f3616707d790e9feae035582d73fce..6b975ec7c5a6da0fb78c0fa530609ad15e451de1 100644
--- a/test/common/matrix_portion.cpp
+++ b/test/common/matrix_portion.cpp
@@ -9,7 +9,7 @@
 #include <algorithm>
 #include <cstddef>
 
-#include "src/kai_common.h"
+#include "kai/kai_common.h"
 #include "test/common/rect.hpp"
 #include "test/reference/round.hpp"
 
diff --git a/test/common/printer.cpp b/test/common/printer.cpp
index 320f7244823e7f17b740834052d7e1b64302194d..7d7144d836bcad603a3a6fdf8c1b37cc6c8f1a6a 100644
--- a/test/common/printer.cpp
+++ b/test/common/printer.cpp
@@ -9,7 +9,7 @@
 #include <ostream>
 #include <string_view>
 
-#include "src/kai_common.h"
+#include "kai/kai_common.h"
 #include "test/common/bfloat16.hpp"
 #include "test/common/data_format.hpp"
 #include "test/common/data_type.hpp"
diff --git a/test/reference/binary_elementwise.cpp b/test/reference/binary_elementwise.cpp
index c5c61a6679433f949dba0b001719c997998e2f40..d5abd26cc18422ee78334b0a63d3f9f9a907c076 100644
--- a/test/reference/binary_elementwise.cpp
+++ b/test/reference/binary_elementwise.cpp
@@ -11,7 +11,7 @@
 #include <cstdint>
 #include <vector>
 
-#include "src/kai_common.h"
+#include "kai/kai_common.h"
 #include "test/common/data_type.hpp"
 #include "test/common/float16.hpp"
 #include "test/common/int4.hpp"
diff --git a/test/reference/cast.cpp b/test/reference/cast.cpp
index 1eae56df1fe556862200aaed4c7105784c106dd2..21486c053e08e1bdc972d3b75f8ffd170ead97f5 100644
--- a/test/reference/cast.cpp
+++ b/test/reference/cast.cpp
@@ -8,7 +8,7 @@
 #include <cstdint>
 #include <vector>
 
-#include "kai_common.h"
+#include "kai/kai_common.h"
 #include "test/common/bfloat16.hpp"
 #include "test/common/data_type.hpp"
 #include "test/common/memory.hpp"
diff --git a/test/reference/fill.cpp b/test/reference/fill.cpp
index 238397c0f88816aa7bc98552b1a9c6a78aa34dc4..d58ab6ce6f7ef87a833b19afadd910442c981580 100644
--- a/test/reference/fill.cpp
+++ b/test/reference/fill.cpp
@@ -13,7 +13,7 @@
 #include <type_traits>
 #include <vector>
 
-#include "src/kai_common.h"
+#include "kai/kai_common.h"
 #include "test/common/bfloat16.hpp"
 #include "test/common/data_format.hpp"
 #include "test/common/data_type.hpp"
diff --git a/test/reference/matmul.cpp b/test/reference/matmul.cpp
index 803fa23b6d79b31be8ddd18e64d1b4e3923ae788..984982a2b0b00e0e976f427e488f5ce3df77ea97 100644
--- a/test/reference/matmul.cpp
+++ b/test/reference/matmul.cpp
@@ -10,7 +10,7 @@
 #include <cstdint>
 #include <vector>
 
-#include "src/kai_common.h"
+#include "kai/kai_common.h"
 #include "test/common/data_format.hpp"
 #include "test/common/data_type.hpp"
 #include "test/common/float16.hpp"
diff --git a/test/reference/pack.cpp b/test/reference/pack.cpp
index a61efff84fe11fd9a625baa1af2269bf9c869aff..261d7a53cdaf0e614285263ec3b97c0283521e8f 100644
--- a/test/reference/pack.cpp
+++ b/test/reference/pack.cpp
@@ -14,7 +14,7 @@
 #include <limits>
 #include <vector>
 
-#include "src/kai_common.h"
+#include "kai/kai_common.h"
 #include "test/common/data_format.hpp"
 #include "test/common/data_type.hpp"
 #include "test/reference/quantize.hpp"
diff --git a/test/reference/quantize.cpp b/test/reference/quantize.cpp
index 141892b2b7d1befd6642e4b836caf45d93c9e57b..213328d5065cf4b1c24f20f94fb7b794bae908d1 100644
--- a/test/reference/quantize.cpp
+++ b/test/reference/quantize.cpp
@@ -13,7 +13,7 @@
 #include <tuple>
 #include <vector>
 
-#include "src/kai_common.h"
+#include "kai/kai_common.h"
 #include "test/common/data_type.hpp"
 #include "test/common/int4.hpp"
 #include "test/common/memory.hpp"
diff --git a/test/reference/reduce.cpp b/test/reference/reduce.cpp
index dff4cf696cd1a1b6ea38383fc9ec590759c06b57..0e83b9bfdc7cf226635100e03cdd84c19497e207 100644
--- a/test/reference/reduce.cpp
+++ b/test/reference/reduce.cpp
@@ -10,7 +10,7 @@
 #include <cstdint>
 #include <vector>
 
-#include "src/kai_common.h"
+#include "kai/kai_common.h"
 #include "test/common/data_format.hpp"
 #include "test/common/data_type.hpp"
 #include "test/common/int4.hpp"
diff --git a/test/reference/transpose.cpp b/test/reference/transpose.cpp
index bee2182a120e46f9df3d25832a2f29e6c9c16849..95cbc8e22566888fc13e7353abc6b35d5e837425 100644
--- a/test/reference/transpose.cpp
+++ b/test/reference/transpose.cpp
@@ -11,7 +11,7 @@
 #include <cstring>
 #include <vector>
 
-#include "kai_common.h"
+#include "kai/kai_common.h"
 #include "test/common/data_type.hpp"
 
 namespace kai::test {
diff --git a/test/tests/matmul_test.cpp b/test/tests/matmul_test.cpp
index 544d17e93046d63a9d6ece2236cc5d73909eca5f..65c5f60344a7a87fa2b5a4ae6bea27cbdc1db84e 100644
--- a/test/tests/matmul_test.cpp
+++ b/test/tests/matmul_test.cpp
@@ -20,9 +20,9 @@
 #include <utility>
 #include <vector>
 
-#include "src/kai_common.h"
-#include "src/matmul/kai_rhs_pack_kxn_f16p16x1biasf16_f16_f16_neon.h"
-#include "src/matmul/matmul_clamp_f16_f16_f16p/kai_matmul_clamp_f16_f16_f16p16x1biasf16_6x16x8_neon_mla.h"
+#include "kai/kai_common.h"
+#include "kai/ukernels/matmul/matmul_clamp_f16_f16_f16p/kai_matmul_clamp_f16_f16_f16p16x1biasf16_6x16x8_neon_mla.h"
+#include "kai/ukernels/matmul/pack/kai_rhs_pack_kxn_f16p16x1biasf16_f16_f16_neon.h"
 #include "test/common/compare.hpp"
 #include "test/common/data_format.hpp"
 #include "test/common/data_type.hpp"