diff --git a/.github/workflows/build-run-kernel-snake.yml b/.github/workflows/build-run-kernel-snake.yml
index b0dc2cea..e569a167 100644
--- a/.github/workflows/build-run-kernel-snake.yml
+++ b/.github/workflows/build-run-kernel-snake.yml
@@ -20,4 +20,4 @@ jobs:
         working-directory: kernels/${{ matrix.kernel }}
     strategy:
       matrix:
-        kernel: [alloc, simple_copy, transform_copy]
+        kernel: [alloc, simple_copy, transform_copy, gemm]
diff --git a/.github/workflows/build-run-kernel.yml b/.github/workflows/build-run-kernel.yml
index a5b569c0..c547249f 100644
--- a/.github/workflows/build-run-kernel.yml
+++ b/.github/workflows/build-run-kernel.yml
@@ -20,4 +20,4 @@ jobs:
         working-directory: kernels/${{ matrix.kernel }}
     strategy:
       matrix:
-        kernel: [streamer_alu, tiled_add, streamer_matmul, gemmini, rescale, gemm]
+        kernel: [streamer_alu, tiled_add, streamer_matmul, gemmini, rescale]
diff --git a/kernels/gemm/Makefile b/kernels/gemm/Makefile
deleted file mode 100644
index fc6c0b58..00000000
--- a/kernels/gemm/Makefile
+++ /dev/null
@@ -1,43 +0,0 @@
-# Courtesy of Federico Ficarelli
-
-.DEFAULT_GOAL := all
-
-include ../../runtime/snax-gemmx.rules
-include ../../runtime/Makefile.rules
-
-TESTS =
-TESTS += gemm.x
-
-MLIRPREPROCFLAGS = --linalg-generalize-named-ops
-MLIRPREPROCFLAGS += --mlir-print-op-generic
-MLIRPREPROCFLAGS += --mlir-print-local-scope
-
-%.preprocfinal.mlir: %.mlir
-	$(MLIROPT) $(MLIRPREPROCFLAGS) -o $@ $<
-
-
-SNAXOPTFLAGS = -p convert-linalg-to-kernel,insert-accfg-op{accelerator=snax_gemmx},dispatch-kernels,convert-linalg-to-stream,fuse-streaming-regions,snax-bufferize,alloc-to-global,set-memory-space,set-memory-layout,realize-memref-casts,insert-sync-barrier,dispatch-regions{nb_cores=2},convert-stream-to-snax-stream,convert-linalg-to-accfg,convert-accfg-to-csr,snax-copy-to-dma,memref-to-snax,snax-to-func,clear-memory-space
-
-CFLAGS += -std=gnu11
-CFLAGS += -Wall -Wextra
-
-data.c data.h:
-	$(PYTHON) gendata.py
-
-%.x: %.o main.o data.o
-	$(LD) $(LDFLAGS) $^ -o $@
-
-sim_%: %
-	rm -fr ./logs/
-	$(VLTSIM) $<
-
-RUN = $(addprefix run_, $(TESTS))
-$(RUN): run_%: sim_%
-	mv logs $(subst sim_,,$<).logs
-
-all: $(TESTS)
-
-allrun: $(RUN)
-
-clean:
-	rm -fr *.ll12 *.x *.o *.logs/ logs/ data.h data.c
diff --git a/kernels/gemm/Snakefile b/kernels/gemm/Snakefile
new file mode 100644
index 00000000..cd2ef795
--- /dev/null
+++ b/kernels/gemm/Snakefile
@@ -0,0 +1,80 @@
+from util.snake.configs import get_snax_gemmx_config
+
+config = get_snax_gemmx_config()
+config["mlirpreprocflags"] = [
+    "--linalg-generalize-named-ops",
+    "--mlir-print-op-generic",
+    "--mlir-print-local-scope",
+]
+config["snaxoptflags"] = ",".join(
+    [
+        "convert-linalg-to-kernel",
+        "insert-accfg-op{accelerator=snax_gemmx}",
+        "dispatch-kernels",
+        "convert-linalg-to-stream",
+        "fuse-streaming-regions",
+        "snax-bufferize",
+        "alloc-to-global",
+        "set-memory-space",
+        "set-memory-layout",
+        "realize-memref-casts",
+        "insert-sync-barrier",
+        "dispatch-regions{nb_cores=2}",
+        "convert-stream-to-snax-stream",
+        "convert-linalg-to-accfg",
+        "convert-accfg-to-csr",
+        "snax-copy-to-dma",
+        "memref-to-snax",
+        "snax-to-func",
+        "clear-memory-space",
+    ]
+)
+
+
+module default_rules:
+    snakefile:
+        "../../util/snake/default_rules.smk"
+    config:
+        config
+
+
+use rule * from default_rules as default_*
+
+
+rule compile_main:
+    input:
+        "main.c",
+        "data.h",
+    output:
+        "main.o",
+    shell:
+        "{config[cc]} {config[cflags]} -c {input[0]}"
+
+
+rule all:
+    input:
+        "gemm.x",
+    shell:
+        "{config[vltsim]} {input[0]}"
+
+
+from gendata import create_data_files
+
+
+rule generate_data:
+    output:
+        "data.c",
+        "data.h",
+    run:
+        create_data_files()
+
+
+rule link_snax_binary:
+    input:
+        "gemm.o",
+        "main.o",
+        "data.o",
+    output:
+        "gemm.x",
+    shell:
+        "{config[ld]} {config[ldflags]} {input} -o {output}"
diff --git a/kernels/gemm/gendata.py b/kernels/gemm/gendata.py
index 62ea1633..aa530748 100755
--- a/kernels/gemm/gendata.py
+++ b/kernels/gemm/gendata.py
@@ -1,10 +1,9 @@
-# simple script to generate inputs and expected outputs for simple_matmult
-
 import numpy as np
 
 from util.gendata import create_data, create_header
 
-if __name__ == "__main__":
+
+def create_data_files():
     # Reset random seed for reproducible behavior
 
     np.random.seed(0)
@@ -15,7 +14,6 @@
 
     A_size = [m, k]
     B_size = [k, n]
-    O_size = [m, n]
 
     # D = A.B + C
     low_bound = -128
diff --git a/util/snake/configs.py b/util/snake/configs.py
index 86411ed2..3e7ebb1f 100644
--- a/util/snake/configs.py
+++ b/util/snake/configs.py
@@ -19,3 +19,17 @@ def get_snax_mac_config():
         f"-I{snitch_sw_path}/target/snitch_cluster/sw/snax/mac/build/mac.o"
     )
     return config
+
+
+def get_snax_gemmx_config():
+    # use CONDA_PREFIX to access pixi env
+    snax_utils_path = os.environ["CONDA_PREFIX"] + "/snax-utils"
+    snitch_sw_path = snax_utils_path + "/snax-kul-cluster-mixed-narrow-wide"
+    config = {}
+    config.update(get_default_paths())
+    config.update(get_default_flags(snitch_sw_path))
+    config["vltsim"] = (
+        snax_utils_path
+        + "/snax-kul-cluster-mixed-narrow-wide-rtl/bin/snitch_cluster.vlt"
+    )
+    return config